MEMO Automated Data Capture Project by mercy2beans108

VIEWS: 27 PAGES: 30

									 Can accurate patient-specific dispensed
     prescribing data be captured
    electronically from community
             pharmacies?

                    Grant No: K/OPR/2/2/271

                                  Final Report
                          By MEMO, University of Dundee

        Department of Clinical Pharmacology & Therapeutics
              Ninewells Hospital and Medical School
                         Dundee DD1 9SY


 Dr T M MacDonald
Tel: 01382 632575
Fax: 01382 642637
e-mail: tom@memo.dundee.ac.uk


Technical enquiries to:
Dr M M McGilchrist
Senior Computer Programmer
e-mail: mark@memo.dundee.ac.uk
Executive Summary
Aim
To test the feasibility, accuracy and workload implications of capturing patient-specific dispensed
prescribing data from community pharmacies.

Method
The project had five principle stages:

1.   The selection of 5 suitable pharmacies.
2.   Installation, maintenance and operation of ISDN lines and bridging hardware.
3.   The development of network based on-line services within MEMO.
4.   An evaluation of the economic viability of the intervention within the pharmacy at the level of
     both the system software and staff working practices including user acceptance through
     observations and questionnaires.
5.   An evaluation of the quality of the data captured.

Stage 1
Pharmacies were chosen on the basis of patient population characteristics that might influence the
outcome of the project: transient population or captive population, adverse social status, drug addiction
within the community, dispensing rates, etc. Three city and two rural pharmacies were chosen. The city
centre pharmacy was sold shortly before incorporation into the project so we report only on the
remaining four pharmacies.

Stages 2 & 3
Secure, transparent and dynamic access by pharmacy systems to the MEMO network was provided by
ISDN bridging hardware. A MEMO supplied software module embedded in the pharmacy system
enabled real-time access to two on-line services in MEMO:

    to allocate the CHI
    to receive dispensed prescription records
The operation of the system can be described thus: a patient visiting a pharmacy for the first time was
identified and the CHI number (and associated data) stored locally for maintenance of the patient
medication record (PMR). This made a future call to the central service unnecessary and in this way
ISDN calls to the central site were kept to a minimum. Secondly, at a time chosen by the pharmacist
(usually once per day) the prescription data were submitted to the deposit service.

Stage 4 Time & motion study
Working practice within the pharmacy was influenced as a result of specific changes in the pharmacy
system user interface. These changes were introduced at two points in the procedure that generated a
label for the dispensed medication. A third change was introduced as a utility operation. The
consequences of these changes were quantified through an objective time and motion study and from
the results it is clear that the changes did not affect the overall time to fill a prescription. In addition,
the staff perceptions of the system were positive. The workload and pace of the pharmacy staff did not
have to be altered to obtain data from community pharmacies. This acquisition was cost effective and
captured high quality patient medication data.

Stage 5 Data validation
The data were validated to ensure that the electronic submissions by any pharmacy were accurate. We
compared the electronic data captured in pharmacies with the actual prescriptions dispensed and
submitted to the pricing agency for payment and entered by MEMO. It was established that no software
faults or procedures in the pharmacy had substantially affected the accuracy of the data. A comparison
was conducted of 14,000 items that allowed close inspection of the data where necessary. The results
are summarised below.

Additional items collected by pharmacy                                             9.7%
Inability to allocate a CHI (tourists etc)                                         4.2%
CHIs allocated in error                                                            <1%



                                                     2
Records containing other errors                                            0.1%
Duplicate records submitted                                                0.5%

The high rate of inability to allocate a CHI number was due to 2 of the 5 pharmacies serving a high
proportion of non Scots tourists.

Conclusions
In conclusion, electronic capture of patient specific dispensed prescribing:

   Was technically feasible
   Worked well in practice
   Did not impact significantly on pharmacist time to dispense a prescription.
   Was viewed positively by pharmacists & pharmacy technicians
   Uncovered some issues that can be overcome by technical or procedural
    modifications
   Provided accurate on-line patient-specific dispensed prescribing data

The success of the present project has major implications for the NHS in Scotland.
Implementation of a similar system throughout Scotland would reap significant
rewards in:

 Improved patient compliance
 Improved pharmaceutical care
 Improved drug interaction checking
 Improved prescribing audit and formulary control
 Improved drug utilisation data / trends
 Improved fraud detection
 Improved capture of dispensing data by the Pharmacy Practice Division and
  reduced costs of this
 Improved pharmacovigilance through record-linkage
 Improved outcomes research
 Improved pharmacoeconomic evaluations of new pharmaceuticals




                                                 3
                                       Introduction

If the NHS is to improve patient care and the management of resources, control over the generation and
use of data is essential. The choice of data and the circumstances under which it is generated must be
well specified and well constrained. Without the latter, the use and interpretation of any data will vary.
A crucial data item is the unique patient identifier without which all other data collection is impaired.

An example of an existing impaired data set is that of community dispensed medication maintained by
community pharmacy systems. Here an individual's data may be scattered over more than one system
with no possibility of linkage at the time of dispensing. In addition the pharmacist is forced to turn
paper information into electronic information in a manner which varies from system to system, and is
deprived of useful information which might be considered too sensitive to be placed on the paper
prescription. The net effect of all this is to reduce the role of community pharmacy in primary health
care and to deprive the NHS of a crucial data set for management and research. For example, this data
set would permit automatic prescription pricing by government agencies or the investigation of a
particular drug's efficacy or safety.

This project looks at the possibilities for real-time, on-line network based systems to help in the
generation and onward use of health care data related to community prescribing and thereby
demonstrate the applicability to other health care data. In particular it demonstrates the deployment of
the Scottish Community Health Index (CHI) into community pharmacy over wide area network links,
and the maintenance of a central repository of linked prescribing data.




                                                    4
                  Acronyms and abbreviations

CHAP   Challenge Handshake Authentication Protocol
CHI    Community Health Index
CSC    Computer Sciences Corporation
DES    Data Encryption Standard
PPD    Pharmacy Practice Division
ISDN   Integrated Services Digital Network
JRC    John Richardson Computers Ltd.
LAN    Local Area Network
OTC    „Over the counter‟
PAS    Patient Administration System
PMR    Patient Medication Record
SOP    Standard Operating Procedures
VPN    Virtual Private Network
WAN    Wide Area Network




                                              5
                                        Background

MEMO is a university based research organisation specialising in pharmacoepidemiology,
pharmacoeconomics, outcomes research and disease registers. Acquisition of accurate data is of prime
importance to such research. With the exception of electronic hospital discharge summaries, laboratory
data and the General Registrar's Office death certification data, all electronic data are created by
MEMO from paper records with the assistance of data entry clerks. All these data sets are record linked
through the Community Health Index – an index of all patients registered with Scottish NHS GPs that
provides a unique patient identifier.

An important data set is that of all community dispensed medication in the Tayside region of Scotland
(population 400,000). MEMO links these data with other data sets to perform record-linkage studies of
drug safety, etc. Paper prescriptions are received from the pricing authority – the Pharmacy Practice
Division - and entered into a computer database. An in-house demographic database (created and
maintained from the official CHI) is used to accurately (~98.5%) ascribe the patients‟ CHI number to
each prescription. This operation is costly and expansion to a Scottish-wide data set is infeasible by this
method. To reduce costs and enable Scottish-wide data MEMO has explored automated means of
capturing these data.

MEMO seeks to capture patient-specific community dispensed prescribing information electronically
using local area networks (LANs) linked by the ISDN - a wide area network (WAN). This project
involved the use of DOS based pharmacy systems to test the feasibility of such an approach:
technically, operationally and economically.




                                                    6
                                     Project Details

The project had five main stages:

1.   The selection of 5 suitable pharmacies within Tayside. These pharmacies all ran software from
     John Richardson Computers Ltd, chosen for its wide market penetration at the time.
2.   Installation, maintenance and operation of ISDN lines and bridging hardware. British Telecom was
     consulted on the choice of bridging hardware capable of supporting the overall software
     application, particularly regarding service response times at the pharmacies and management of
     network traffic to minimising call costs.
3.   The development of real-time network-based services within MEMO. Specifically, the provision
     of two services at the network level: a service for tracing patient CHI numbers and a service for
     depositing prescription data centrally within MEMO. An intervention within the pharmacy at the
     level of both the system software and staff working practices. A software module provided by
     MEMO was added to the JRC pharmacy system software to provide a simple and efficient
     mechanism for accessing MEMO services. JRC modified the pharmacy system software in three
     specific areas to provide user access to these services. (1) When attempting to identify patients
     locally, a new patient would have his or her PMR defined by the CHI tracing service. (2)
     Additional data was collected over and above what was required for generating the medication
     label, and (3) a simple utility was provided to allow pharmacy staff to submit dispensed
     prescribing data for central storage at MEMO at a time of their own choosing.
4.   An evaluation of the economic viability of the system was done through a time and motion study
     comparing times before intervention to those 3 months after intervention. In addition, acceptance
     of the system by users was assessed through observation and questionnaires both before and after
     intervention.

5.   The completeness and quality of the data captured was evaluated carefully by comparison with the
     same prescription data delivered by PPD and entered by MEMO staff as part of MEMO's normal
     prescription data entry.




                                                   7
                                      Pharmacy Selection

Stage 1
Pharmacies were available for selection if the following criteria were satisfied:

   The pharmacies had to be located in Tayside since MEMO receives paper prescriptions from the
    PPD for all of Tayside as a matter of routine.
   The pharmacies had to run the DOS single user version of JRC's pharmacy system.
   The pharmacies had to have patient population characteristics that might influence the outcome of
    the project: transient population, captive population, adverse social status, drug addiction within
    the community, etc.
   The pharmacies had to be not too far from the nearest ISDN exchange if line conditions were to be
    guaranteed satisfactory.

Five pharmacies were chosen to be representative of all types of pharmacy. In addition we chose two
pharmacies based on 'worst case' locations where we expected high numbers of tourists. The five
pharmacies were designated:

   Small rural
   Small semi-rural
   Medium semi-rural
   Medium city centre
   Large city centre

Unfortunately, the large city centre pharmacy was sold shortly before incorporation into the project
(but after installation of the ISDN lines). The new owner installed their own software into the shop and
could not participate further with the project. This was beyond our control. The project therefore
reports on the four pharmacies that remained.




                                                    8
                                                 ISDN

Stage 2
At the inception of the project, MEMO staff had little working knowledge of ISDN and the required
network bridging hardware. It was therefore decided to use British Telecom for hardware procurement
and installation.

The following ISDN lines were installed by BT:

1 basic rate ISDN line (2 channels x 64kb/s) in each pharmacy (5)
1 basic rate ISDN line in MEMO for internal use
1 primary rate ISDN line (30 channels x 64kb/s) in MEMO for service provision

The bridging hardware was required to bridge the IPX LAN protocol and provide protocol spoofing
and toll saving facilities. In late 1995 such requirements meant expensive bridges. BT recommended
KNX Remote Access POD 2100 bridges in the pharmacies and a KNX-IS Primary Rate Access Server
(hosted by an i486 PC) within MEMO. KNX staff provided an initial bridge configuration and one
day's training to MEMO technical staff. During the period of the project MEMO technical staff became
very experienced in the use of ISDN for LAN-LAN bridging.

The bridging hardware provided point-to-point authentication through challenge handshake
authentication protocol (CHAP). Since the remote sites (pharmacies) would only host a single PC
bridge authentication was considered sufficient access control for MEMO services. The bridges also
provided 112-bit DES encryption enabling a virtual private network (VPN). MEMO's client module
and services software did not, therefore, offer authentication and encryption.

The pods allowed remote access and maintenance of their configuration information and, along with
bridge management software, dynamic management of the ISDN WAN links was possible. This was
considered essential for a system of this kind.

A large number of products are now available providing the same functionality and at much lower cost.




                                                  9
                                         MEMO Services

Stage 2
The MEMO PC servers ran three network services throughout the project: a CHI tracing service, a
prescription deposit service and an exchange service. The latter broadcasts its presence on the MEMO
LAN (and over the ISDN WAN links) to the MEMO client module integrated into the pharmacy
systems. The client module locates other services using the exchange service making the entire process
dynamic.

These services have proved to be highly stable, a result of simple and efficient designs. These services
have never failed despite nearly two years of continuous use. Thus, in the present system, scheduled
loss of service takes place only when a new CHI database is installed. This occurs twice per year and
takes just 20 minutes on each occasion, thus providing 24x7 availability at the 99.9924% level.

CHI Tracing Service

The present MEMO CHI database is static (for both records and indexes). Twice per year, CSC
provides the entire Scottish CHI, consisting of 6 files for 8 consortia, on industry standard 9-track tape.
These flat record dumps are re-built and extensively indexed for each original file. The indexes consist
of inverted lists for each unique symbol (see later) found in the records of each file. Each symbol in
each record is a symbol instance and the largest file, containing over 2M records, has ~35M symbol
instances indexed. Over the entire Scottish CHI ~100M symbol instances are indexed providing
comprehensive ability to trace CHI numbers. Performance figures are equally impressive with a typical
trace (e.g. 17 John White) taking under 0.2s on a 120MHz Pentium with 32Mb RAM and a SCSI disk
subsystem. CHI translation typically takes less than 0.03s.

A fully dynamic CHI could have been provided without sacrificing the index design. Although there
was a need for such a facility no feasible method could be found by staff at CSC to present MEMO
with updates at regular time intervals. The official CHI is defective in this respect: there is no way to
definitively identify updated CHI records from their field contents or other database information.

CHI updates take place by various routes:

   Patient submission of changes of address to general practice
   Patient re-registration with general practice
   Hospital PAS systems

Methods of intercepting these updates are being investigated and a future MEMO CHI will be fully
dynamic. However, for this project an update resolution of 6 months was provided.

Prescription Deposit Service

The prescription deposit service functioned like a file transfer protocol with additional consistency
checks. Prescription item records created on the pharmacy system since the last deposit were grouped
and transferred by the MEMO client module for central storage by the service. An audit record was
created for each group deposited. Each deposit had to be complete and partial depositions (through
network failure, etc.) were cancelled. These could be re-tried at any time.

The transfer consisted of a header record, one or more prescription item records and a trailer record.

The header record contained details of the pharmacy, a MEMO assigned pharmacy identifier used by
all services, and deposit details such as the time the prescription items were initially grouped for
transfer. The latter may not match the actual time of transfer if previous transfers had failed. Under
such conditions multiple groups would accumulate for transfer.

The trailer record contained only the number of prescription item records in the group. The deposit
service cancelled the transfer should this count not match the actual number of items in the transfer.




                                                    10
All outstanding groups were transferred at the same time and the pharmacist was not required to have
any knowledge about individual groups.

Prescription item records came in two types: (1) a new item designated „D‟ not submitted to the service
before and (2) an item designated „E‟ that should match a previous 'D' item submitted to the service and
which acted to cancel that previous item. For example, when a pharmacist edited a PMR entry to
produce a corrected label an ED item pair was produced to accompany the original D item. MEMO
software that processed the contents of the prescription repository recognised such DED item
sequences (and any higher order sequences).

Prescription item records consisted of a number of ASCII text fields and provided the following
information:

   Type - NHS, dental, nurse, hospital, OTC, private
   Date and time of dispensing
   Patient's CHI number
   GP code from GP10 form
   Code for preparation prescribed
   Description of preparation prescribed (as predefined text)
   Code for preparation dispensed
   Description of preparation dispensed (as predefined text)
   Amount dispensed as number and unit
   Dosage (as predefined text)
   Additional directions (free text)

No agreed system of coding preparations is used by commercial pharmacy systems and no guarantee is
made that a code will be used. As a consequence preparations could only be identified by their textual
description. However, different systems have subtly different descriptions for the same preparation and
many systems permit the pharmacist to define local descriptions for preparations. The same remarks
apply to dosage and amount making processing of the repository without error or without intervention
very difficult. The need for agreed coding standards is clear.




                                                  11
                                        Pharmacy Intervention

Stage 3
The original procedure for preparing a label for dispensed medication was as follows.

Stage 1. The patient was identified on the pharmacy system usually by use of the surname and an
initial. A small pick list was presented and the pharmacist or technician would choose the appropriate
patient record. If the patient was not already present on the system the pharmacist was required to
create a new PMR by entering all the relevant details: name, title, address, postcode and prescribing GP
(not necessarily the registered GP). Date of birth and allergy information might also be entered if these
were available. Only once this procedure was complete (or else avoided) could the pharmacist return to
producing a label.

Stage 2. Once the patient was identified the pharmacist enters the details of each item on the
prescription requiring a label. This would involve selection of the required preparation, the amount to
be dispensed and directions for use.

Stage 3. At this point the label was printed, the medication record saved to disk and any stock control
decisions taken.

Stages 2 and 3 were be repeated for each item on the prescription. In this way PMR data accumulated
at the local pharmacy but was not be directly available to any other health care professional, manager
or researcher.

As already stated, the pharmacy system software was augmented in three specific areas to modify this
procedure and to enhance the value of the PMR.

When a patient was not identified locally, or had a PMR without a defined CHI number, the pharmacy
system offered the use of the MEMO CHI tracing service to identify the patient and define the PMR.
This could be done within the normal flow of label production and did not require use of specific
utilities elsewhere in the software. The CHI tracing service identified patients by an associative match
of record lists maintained for each syntactic element of a patient's demographics, this syntactic element
being known as a symbol. For example, in the following CHI record definition:

Name              Address                     Post-Code        Reg. GP   DOB                  CHI
John McIver       67 Seafield Ave, Dundee     DD2 2SD          T7774     12/08/1898     1208980072


the following symbols were defined:

JOHN, MCIVER, *IVER, 67, SEAFIELD, DUNDEE, DD22SD, DD22, T7774, T1206, 12/08/98

Of special note here are the symbols *IVER which allows for seamless MC/MAC lookups and T1206
which allows for practice lookups should the prescribing GP not be the registered GP. Any
combination of the above symbols can be used to identify the patient and usually generates a small pick
list from which the patient can be selected. Because there are many (combinatorial) ways to identify the
patient, spelling errors and mistakes are easily circumvented, both on the prescription and in the
database.

When the pharmacist chose to create a PMR for a new patient he was presented with an edit line in
which to type some symbols. After selecting the required patient from the pick list the CHI tracing
service automatically pre-loads the PMR record with the following details:

CHI number, title, gender, name, address, postcode, registered GP code, GP name, registered practice
code, address and telephone number.

If the pharmacist was satisfied with the PMR presented he could save the record and continue with
label production.

Dummy CHI numbers were provided for exceptional conditions, these being:



                                                    12
    Temporary residents from outside Scotland
    Patients who could not be located on the database but should have been, given sufficient
     information was available
    Patients who could not be identified since the prescription was difficult to read and the patient, or a
     representative, was not present to provide additional information
    Veterinary prescriptions

It should be noted that each patient so designated was still given a unique identifier by the local
pharmacy system and the pharmacist still entered PMR details manually. This allowed for correct
allocation of the CHI number at some point in the future. However, any of the patients' data transmitted
at this stage could not be identified.

Once allocated, the CHI number need never change in the PMR of the pharmacy system. Nevertheless,
issues such as synchronisation of patient details in the CHI service and pharmacy PMR are important.
Whenever a patient was identified by the local pharmacy system (by surname and initial) their current
PMR details were always presented in the resulting pick list. The pharmacist was therefore offered
ample opportunity to note any mismatch between the PMR details and those on the prescription. At this
point, as with a new patient, the pharmacist could use the CHI tracing service to redefine the PMR
contents. No better methods were employed to maintain PMR and CHI service synchronisation
although many options were available. For example, the system could have employed timeouts which
forced re-use of the CHI tracing service to verify PMR contents, or perhaps 'behind the scenes'
verification could have been used as there was substantial unused bandwidth with the ISDN links.

A virgin system, where local PMRs contain no CHI numbers, would incur high call costs from the
inevitable CHI tracing. Ultimately a steady state would arise in allocating CHI numbers. However, to
avoid this unnecessary cost an automatic CHI allocation procedure was adopted immediately prior to
use of the augmented pharmacy system. Existing patient details were exported from the pharmacy
system and patients identified by probability matching using software within MEMO. For those PMRs
with good matches the new record details were then re-imported with an attached „flag‟ to indicate that
the new patient details required verification. When a patient presented with a prescription the normal
local lookup procedure would be followed. However, when an existing patient was selected, the
patient's details would automatically be offered for verification. The pharmacist could then save the
record as is, clearing the verification flag, or could decide to create a new PMR after due consideration.
In this way patients with verified CHI numbers are never prompted and the system proceeds directly to
stage 2.

Directly before stage two we sought to collect other useful information not normally available. In
particular we wished to collect the prescribing GP code and the declared exemption status for each
prescription. This involved the insertion of an edit line for the collection of the GP code, and the
selection of the exemption status letter (on the reverse side of the GP10) from a menu, prior to defining
the label for the first item. The pharmacist did not actually need to type the GP code as a practice pick
list was directly available (from the CHI tracing service). It is important to note that this data collection
did not require the direct use of the CHI tracing service and the consequent call costs since the
pharmacy system was designed to cache GP and practice details. These caches were updated whenever
the central CHI tracing service was updated (see above).

The intervention in the pharmacy can be summarised simply:

1.   An additional in-line procedure for new patients presenting at the pharmacy.
2.   The collection of additional data (GP code and exemption status) prior to defining the label for the
     first prescription item.
3.   A new utility that allowed for the transmission to MEMO of all data recorded since the last
     transmission.

For a pharmacist already committed to maintaining accurate PMRs the first procedure saved time since
no manual typing of the PMR details is required. The collection of the additional data, however, cost
time since this data was not previously being collected. The prescription transfer had no time
implications since the once daily transmission took place at ~1000 prescription items per minute over
the ISDN line.




                                                     13
While the prescription transfer process is trivial for the pharmacist there were a number of issues that
had to be addressed relating to data integrity. The central prescription repository had to be kept
synchronised with the pharmacy system PMRs. Consider that a pharmacist is free to edit an item after
transmission or only partially dispense an item prior to transmission (owings). In cases of the former,
subsequent transmissions would involve pairs of items – an „E‟ item cancelling the previous item
transmitted and a D item defining the new contents. MEMO software that processes the contents of the
prescription repository had to recognise such DED item sequences (and any higher order sequences). In
the case of an owing the individual records transmitted would sum in amount to the original item. In
this project no record structures were defined to link such owings explicitly although clearly such
linkage is necessary.




                                                  14
                            Technical development

                                            Introduction


MEMO has had a CHI tracing service since mid-1990 for use by data entry staff to provide patient-
specific dispensed prescription information from paper GP10s. Such a system is costly and the need for
an automated system for the collection of this data was apparent from the beginning. An obvious
answer was to ascribe the CHI at the pharmacy. However the issue has always been how to provide a
CHI tracing service to community pharmacy.

In late 1992 a project was run involving a MEMO pharmacy system alongside a pharmacy system from
an existing supplier. The MEMO system lacked stock ordering, which prevented its direct use, but it
did provide for CHI tracing. A copy of the CHI for Tayside was held on the pharmacy machine and a
pharmacy student was employed to enter the data in parallel with the pharmacist. This project showed
that accurate patient-specific information could be collected from pharmacies but raised the obvious
issue of maintaining the CHI locally. Providing copies of the CHI on CD-ROM on a monthly basis was
suggested but would only have raised new issues, one of which being security.

In mid-1993 ISDN routers and bridges became widely available. This offered the prospect of providing
the CHI to community pharmacies directly from MEMO as the ISDN hardware enabled the extension
of the internal MEMO CHI system into any pharmacy. To facilitate this, client/server applications with
low bandwidth requirements were developed to enable this service on pharmacy computer systems.
John Richardson Computing (JRC) is a major pharmacy software supplier and agreed to become
involved with the pilot project.

The project started in August 1995 and involved the development of both a CHI tracing service and
prescription deposit service.

                                                  ISDN


The ISDN is a public digital network available world wide for the fast transmission of any digital data.
ISDN is becoming more common in the work place and at home. Digital connections between two
points on the network can be made and broken almost instantaneously permitting pay-as-you-use. It is
possible to connect two or more local area networks (LANs) into a single LAN using ISDN links and
associated technology. Connections between points on the network separated by ISDN links are logical
in the sense that the ISDN links are only in use (and charged for) during transmission of data between
these points. It is possible to install any desired bandwidth (data transfer rate) at a particular site from
cheap low bandwidth basic rate ISDN2 connections up to expensive high bandwidth primary rate
ISDN30 connections.


                              MEMO-Pharmacy network links


Remote access was achieved through ISDN bridging hardware at both the remote site and central site.
These bridges provided secure, transparent access to the central LAN through a number of features.
Dial-on-demand provided good application response times (< 0.5s) and auto-disconnect criteria kept
line use to a minimum. In combination with hardware security features a low cost virtual private
network (VPN) was provided. The network topology is shown in figure 1. For better economy only a




                                                    15
single B channel was used at each remote site, the other channel then being available for such things as
credit card authorisation.




                                             MEMO
                    CHI Service                                       Deposit Service
                                         MEMO Exchange




                                 Virtual Private Network

                                     MEMO client module
                                                                                 Pharmacy
                                                                                 Computer
                            Pharmacy System Application


Figure 1. Network Topology for MEMO Automated Data Capture Project.



                                 Client-Server Applications


A thin client module was added to the JRC pharmacy system to enable real-time communication with
the central services - CHI tracing and prescription deposit. All procedures initiated by the (MEMO)
client software were performed by these services as one or more primitive operations or transactions,
which were then assembled by the client software into useful results. In this pilot only limited software
authentication mechanisms were employed between client module and services; there was no
authentication of users. The client-server architecture and system topology is shown in Figure 2 and a
summary of system components in Table 1.


Table 1. Summary of system components.

                                 Clients                                        Service
Platform                         Intel based PC                                 Intel based PC
Operating System                 DOS v6.2 or above                              NetWare v3.12
Application                      JRC system (real mode)                         MEMO 32-bit NLM
Comm. Protocol                                                 IPX datagram
Network                                                        Ethernet 802.3




                                                   16
                   CHI Service                               Deposit Service
                                         MEMO

               Ethernet LAN                      ISDN30 bridges - PRI Server
     Private                                     (up to 240 channels x 64kb/s)

     Public


                       Integrated Services Digital Network

     Public

     Private
                                                                    ISDN2 bridge - BRI Server
                                                                    (up to 2 channels x 64kb/s)
       Ethernet LAN

                              Pharmacy              Pharmacy                   Workstation




Figure 2. Client-server architecture and system topology.

The two parts of the system can be described thus: a patient visiting a pharmacy for the first time would
be identified and the CHI (and associated data) stored locally for maintenance of the PMR, so making a
future call to the central service unnecessary. In this way ISDN calls to the central site are kept to a
minimum. Secondly, at a time chosen by the pharmacist (usually once per day) the prescription data
were submitted to the deposit service. Thousands of prescriptions can be downloaded within the typical
minimum line time of 60seconds.


                                     System Development


Software development within MEMO took ~8 months coding. Development proceeded as expected and
no issues arose. JRC integrated the MEMO communications module into their system and the first test
took place in September 1996, one year after the award of the Scottish Office grant.

After a few weeks it became clear that an intermittent fault was present in the communications
although it was not immediately obvious where the fault lay. During this period JRC had been forced
for market reasons to upgrade their system and considerable bugs were introduced into the new
program. Until the bugs were sorted it was difficult for MEMO to accept responsibility for the fault.
Eventually, and after many attempts to find the fault, a full-scale systematic experiment was conducted
within MEMO which showed that the ISDN bridging hardware did not always behave as expected. The
manufacturer and BT were not able to provide an answer for this behaviour (letter enclosed). The exact
fault was never fully traced but a small software „fix‟ was devised that would guarantee the correct
operation of the bridge. Since this fix was introduced the system has run flawlessly and has not
required any intervention on the part of any MEMO staff.




                                                   17
                                    Communication Costs


Running costs

A worst case analysis of the service transaction records indicates a lookup call cost of ~1.1p per
prescription (for rural pharmacies with a high number of temporary residents and long distance call
costs). If ISDN line rental is included (the only other running cost) a total lookup cost of ~1.4p is
obtained for the remote site. A number of ISDN pricing regimes are available with various rates for
installation, line rental and call allowance. The data suggest that call costs can be contained within the
call allowance of some pricing regimes reducing the remote running costs to a fixed quarterly rental
yielding ~0.9p per prescription.

Capital Costs

The capital outlay for each remote site involved a one time cost of ~£400 for installation of a basic rate
ISDN line and ~£1100 for ISDN bridging hardware. Central site capital costs were significant in the
context of this pilot. On a larger implementation this would not be the case. ISDN installation can now
be done for ~£300 and bridging hardware can be purchased from many suppliers for ~£400 although
we have not tested the latter for software compatibility.

                                           Conclusions


In terms of software development the project was very successful. However, the ISDN bridge
behaviour indicated the need for rigorous testing of compatibility between hardware and software
applications.

                                     Future Developments


For full operational use a system needs to be dynamic, scaleable and fault tolerant with software based
security and authentication. Although satisfactory in the present project, the IPX protocol would not be
suitable for wide area networks generally and the IP protocol is recommended for future systems. Web
technology would also confer significant advantages for software distribution, system extensibility and
maintenance. These directions would need to be followed to provide a powerful Scottish wide system.




                                                   18
    Staff training and observations of adherence to
                         training

                                               Training


The ability of pharmacy staff to identify the CHI number of each patient presenting a prescription in a
pharmacy was a key element in this project. It must be understood that, in this project at least, a (non-
dummy) CHI number was allocated to a PMR once and only once. Any error in the allocation of a CHI
produced a systematic error in prescription item data from that pharmacy. Given that pharmacy staff
had access to the patient, their representative or other carers when allocating the CHI number this error
rate was expected to be very small and limited by database irregularities. When difficulties arose
pharmacy staff were to use dummy CHI numbers which would allow for correction at a later stage and
make current data anonymous and safe.

A training package was designed which included a separate manual detailing all the additional features
that were available using the MEMO computer software. This was designed to support a formal
training lecture and tutorial.

The training manual complemented the existing JRC handbook that each pharmacy already used. The
extra manual gave clear and concise instructions on each additional feature. This was thought essential
to assist any locum staff who may be called upon to work in the pharmacy at short notice. Locum
pharmacists have to be familiar with a number of very different systems and are required, often, to
dispense using software having had no formal training on that software.

The pharmacists and dispensing staff were invited to a seminar (in MEMO) on the project and a formal
description and demonstration of the new features. Afterwards, each of the pharmacy staff took a turn
in using the computer system that had been set up to replicate the actual system being installed in their
pharmacies. This included the use of sample prescriptions from their own pharmacy to engender
confidence.

A questionnaire was designed to test the confidence of the pharmacy staff in using the system and
circulated to all the staff at the end of the evening. 7 pharmacists and 6 assistant staff answered all the
following questions identically:

    Has the purpose of the project been fully explained? (Yes)
    Has the training covered those aspects of the project relating to your own responsibilities? (Yes)
    Are there any other issues requiring clarification? (No)
    Do you feel adequate time was given to the training? (Yes)

All felt confident or very confident in operating the system.

The remaining pharmacy, due to time constraints, was exposed to “on the job” training which consisted
of one of the MEMO pharmacists spending two days in the pharmacy after upgrading of the system and
acting in a teaching capacity and as advisor. The response to the subsequent questionnaire was
identical.

In all the pharmacies one of the MEMO pharmacists spent at least the first whole day and some of the
following days in the actual pharmacy acting as an advisor and observer of the system in action. This
provided moral as well as physical support to the dispensary staff and allowed any queries to be dealt
with instantly.




                                                    19
At all times during the project both a MEMO pharmacist and a member of the MEMO technical staff
were available on call via wide area pagers.



                                    Adherence to training


An observational study was conducted after each system had been in place for a minimum of 3 months
to evaluate the degree to which staff were observing the procedures regarding use of the system. The
MEMO office manager, who is extremely experienced in training and monitoring data entry personnel
spent a day in each pharmacy in an observational study.

Most of the time (~98%) the correct procedures were followed. However, aberrations occurred when a
label was generated in the absence of a prescription. This could happen for the following reasons:

   Prescriptions for nursing home residents and those received by telephone from the general practice
    were usually filled prior to receiving the physical prescription. This often created confusion with
    duplicate dispensing (by different staff) when the prescription became available. This duplication
    would later be noticed and cancelled in the system. Occasionally this was not the case and a
    duplicate PMR record was obtained. Stock control was also affected under these circumstances.
   When labels were found to be in error on medication awaiting collection it was often quicker -
    particularly with locums - to create a new label rather than edit the existing PMR entry.
   With regard to recording of the GP code prescriptions arriving late were often signed by a different
    GP from that on the telephone. This led to conflicts between the computerised prescribing doctor
    data and the doctor data that was eventually recorded at MEMO.
   When temporary resident patients presented a prescription, locums sometimes resorted to label
    generation without a PMR update as this was quicker than defining a new PMR by hand (with
    dummy CHI number).




                                                  20
Economic assessment of pharmacy intervention
Stage 4
As stated in the project definition, working practice within the pharmacy was modified as a result of
specific changes in the pharmacy system user interface. These changes were introduced at two points in
the procedure that generated a label for the dispensed medication. A third change was introduced as an
additional utility operation. The consequences of these changes were quantified through an objective
time and motion study. Subjective staff perceptions were investigated through a pre-introduction
survey and interview, a post-introduction debriefing session and casual questioning. During the project
JRC were forced to significantly modify their system for market reasons. This represented an
uncontrolled and poorly specified additional intervention. It was anticipated that pharmacy staff would
not appreciate the distinction between the MEMO software intervention and changes made by JRC.

Aims

1.   To evaluate the time required for dispensing a prescription pre- and post-introduction of the new
     pharmacy system.
2.   To survey the users on the perceived feasibility of CHI number allocation and the extra data
     capture.


                                    Time-and-motion study


Computers were first introduced as a labelling device that took over from hand written labels or the
typewriter. Later, the computer was used to store pharmacy based patient medication profiles. There
was a progression towards making pharmacy less dependent on physical record keeping. Studies
performed at the introduction of computers to hospital pharmacy found that time was saved in the total
dispensing process, especially in clerical and costing functions. Within community pharmacy, time-
and-motion studies of the impact of computer applications have not been extensively studied.

A time study is the measurement of work. It was defined by R.M. Barnes (1961) as analysis used to
determine accurately the time required by a qualified and properly trained person, working at a normal
pace, to do a task. In this project the time analysis determined the efficiencies of the extra data capture
and CHI number allocation in relation to other components of the work of filling a prescription in the
pharmacy. The task under study was dispensing a prescription pre- and post-introduction of the new
pharmacy system. The post-introduction timing was approximately 3 months after introduction. This
allowed sufficient time for the staff to become familiar with the new system.

The capture of the prescribing GP and exemption status was required for each prescription and
consistently added time to the dispensing process. CHI number allocation and automatic PMR
definition is quicker and more useful than the former procedure of entering all the patient details by
hand. In a typical pharmacy, PMR definition happens perhaps every 20 prescriptions, and
consequently, any time gains have little impact on the overall time to dispense. The utility operation of
depositing prescriptions with the MEMO service is negligible in terms of time.

A time and motion study was performed in each pharmacy on 3 days of a week (Monday, Tuesday and
Thursday) and for a sample of prescriptions. The day selection and prescription selection accounted
for variations in workload and time of day. Most pharmacies were busiest in the morning and began to
see fewer prescriptions in the afternoon. This was not true for all pharmacies. In one pharmacy the
prescription volume picked up in the late afternoon when people returned home from work. Taking
every third prescription throughout the day ensured a random selection of the workload of the
pharmacy staff.

A MEMO pharmacist (observer) arrived at the opening of the pharmacy and stayed until closing. The
observer recorded every third prescription throughout the day. An electronic stopwatch was used to


                                                    21
time the processing of the prescription to the nearest second. Time started when the prescription was
handed from the patient to a person working in the pharmacy. The prescription was followed through
the entire dispensing process and time intervals were recorded for each component of the prescription
filling procedure:

Patient interaction with staff. This component related specifically to the act of accepting a prescription
from the patient, presenting the medication to the patient, and any counselling by the pharmacist.
Patient counselling was considered if the time was spent on the proper use of medications, answering
questions on side effects and any other prescription-related information exchange.

Computer time. This component was defined as the time from when typing began until a label was
produced for a prescription. Interruptions, if they consisted of the need to procure more supplies are
included, but a phone call, pharmacist-patient information interactions, etc., were not included.

Filling the prescription / checking the prescription. This component was defined as the time required
to obtain drugs from storage areas, transport them to the work area, count out the needed amount, place
a label on the bottle and add auxiliary labels. This time also included a measure of the time to check
the prescription for correctness by the pharmacist.

Total time to fill the prescription. This was the total time from the patient handing the prescription to
the staff until the final bagged medication was given to the patient by the staff. This gives the total
time to fill a prescription assuming flow from one component directly onto the next. Clearly the extent
to which these components overlap will affect interpretation of the figures.

The time in seconds to perform each task was analysed using a statistical package. A series of analysis
of variance (ANOVA) were used to determine significance between the means of each component pre-
and post-introduction. Statistically significant mean differences were based on a p  0.05.

Results

The results of the time and motion study for each pharmacy are shown in Tables 1 to 4.

Table 1. Time and motion analysis for the medium city centre pharmacy

                                                              Mean time  s.d. (s)
             Component
                                    Pre-introduction (n=64)             Post-introduction (n=69)
Patient interaction                 3416                               4029
Computer time                       2719                               3927 a
Filling prescription                6542                               6438
Total end-to-end time               12552                              14271
a
    Significant at p  0.05

Table 2. Time and motion analysis for the small rural pharmacy

             Component                                        Mean time  s.d. (s)
                                    Pre-introduction (n=33)             Post-introduction (n=26)
Patient interaction                 4023                               4640
Computer time                       2919                               2928
Filling prescription                7651                               8173
Total end-to-end time               14657                              15684

Table 3. Time and motion analysis for the small semi-rural pharmacy

                                                              Mean time  s.d. (s)
             Component
                                    Pre-introduction (n=27)             Post-introduction (n=34)
Patient interaction                 4834                               4528
Computer time                       4028                               4531



                                                   22
Filling prescription               7470                               4027 a
Total end-to-end time              15576                              13050
a
    Significant at p  0.05

Table 4. Time and motion analysis for the medium semi-rural pharmacy

             Component                                       Mean time  s.d. (s)
                                   Pre-introduction (n=32)             Post-introduction (n=44)
Patient interaction                2713                               3019
Computer time                      5048                               3933
Filling prescription               4139                               4537
Total end-to-end time              12777                              11457

The city centre pharmacy (Table 1) showed a significant difference in the computer time pre- and post-
introduction. However, the overall time to fill a prescription was not affected.

The small semi-rural pharmacy (Table 3) showed a significant reduction in time to fill a prescription,
but no change in overall time. This reduction in time to fill a prescription may have been due to factors
other than the introduction of the system such as changes in staff, shop procedures and other changes to
JRC‟s computer system.

The small rural pharmacy (Table 2) and the medium size semi-rural pharmacy (Table 4) showed no
significant differences in any component times or the overall time to fill a prescription.

Discussion

The diversity of location and pharmacy size allowed the system to be studied in varying prescription
volumes and pharmacy working paces. It was particularly important to find out if the new system
could be used in busy pharmacies, recording information on visiting patients, with no consequent time
loss.

It was observed that the time and motion study was highly dependent on the pharmacy staff available
over which there was no control. The significant time difference in prescription filling time for the
small semi-rural pharmacy was due to personnel changes in the 3-month interval. A pre-registered
pharmacy student had been working pre-introduction and had graduated at the time of post-introduction
timing. The significant difference in computer time for the medium city centre pharmacy was due to a
locum pharmacist working at a different pace on the new system. This locum worked various days for
the owner and may have performed duties differently than the owner of the pharmacy. In general,
locum pharmacists may work at different pharmacies with different systems and will not have expert
knowledge of any one system.

From the results of the time and motion study it is clear that the extra procedures do not affect the
overall time to fill a prescription and we believe that the workload and pace of the pharmacy staff do
not have to be altered to obtain data from community pharmacies.




                                                   23
                                    Pharmacy staff survey


Prior to the introduction of the new pharmacy system existing staff were interviewed to establish
knowledge of computer systems in general and their own pharmacy system in particular - the previous
version of the JRC system. At the end of the post-introduction period staff perceptions of the success of
the project were obtained through a debriefing session and casual questioning in the shop.

Survey and interview

The survey and interview consisted of a series of questions designed to evaluate staff competence with
computer systems and also staff perceptions of how well their current program satisfied their needs
regarding patient care and shop management. In retrospect these questions may have conveyed the
impression that MEMO was to perform a wholesale improvement in their systems. This was not the
case, and the questions, aside from those on general computer knowledge, should have stuck firmly to
the specific intervention to be made by MEMO.

All dispensary staff had some previous experience with computers, principally the computer in the
pharmacy or one at home. They felt confident about using the existing programs and were happy with
the existing features - stock control, on-line ordering, drug interaction warnings and PMRs. In the
majority of pharmacies the computer was used by all staff. However, in two pharmacies the
pharmacists used the system exclusively. None of the pharmacists interviewed had attended a course in
computing or Information Technology and it was anticipated that few staff in the chosen pharmacies
would be able to judge issues as they arose.

Debriefing session

These sessions found that the new system worked well and its introduction did not negatively affect the
overall working of the pharmacy. The pharmacists were enthusiastic about the patient information that
was available and the potential access to other information such as drug allergies and indication, that
would help them provide better pharmaceutical care, particularly with OTC medication. The
pharmacists perceived that it took longer to fill a prescription post-introduction, but overall they felt
that the information obtained from MEMO outweighed the extra computer interaction to fill a
prescription. This perception had no foundation in the time and motion study.

A number of issues were raised relating directly to the long term purpose of the MEMO project:

   Many pharmacists want more medical information to be in the PMR such as indication, chronic
    medical conditions and hospital discharge summaries. The information presently available is too
    vague to allow for appropriate counselling and the provision of pharmaceutical care. Some
    pharmacies provide patients with a questionnaire to fill in regarding medical history and known
    allergies. Such information could be supplied through the MEMO system. Pharmacists are health
    care professionals and patients will reasonably expect these professionals to have access to such
    data.

   Patients are sometimes given printed information about their recent medication, when requested, to
    take to their doctor‟s surgery or to a local hospital at admission. The MEMO system potentially
    provides a central repository for all community prescribing.

   PMRs were considered important and maintained at a 90-95% level in most shops. Duplicates
    were acknowledged and removed when discovered. The MEMO system provides for full PMR
    maintenance and the elimination of duplicates through the CHI number.

   Drug-drug interactions are not classified as major, moderate or minor. As a consequence this
    feature was often a nuisance. The MEMO system does not address this issue.

   Pharmacies often have professional inquires about a patient‟s medication profile:

    Doctor‟s surgeries - (home visit) prescribing not entered into a patient‟s notes.


                                                   24
District Nurses - to obtain a patient‟s medication history.
Hospital pharmacy and hospital ward - patients could not remember all their medications.

Clearly the MEMO system can provide such information securely and efficiently.




                                             25
                       Validation of captured data
Stage 5
The first pharmacy „went live‟ in September 1996 and the last pharmacy in June 1997. Two pharmacies
were subsequently sold one at the beginning of the project and one after the evaluation had been carried
out. These have remained off-line since then. Between September 1996 and May 1998 a total of about
200,000 prescriptions were received by MEMO. During this same period 55,000 CHI service
transactions took place representing at most 27,500 CHI lookups.

Two of the pharmacies were deliberately chosen in tourist areas. These were known to serve a
significant number of patients from outside the area (Scotland). The ability to identify patients and
maintain PMRs would therefore be impaired. One pharmacy served a considerable number of drug
addicts and, as a rule, medication was prepared in advance of shop opening, affecting typical staff
working practices. Some pharmacies clearly had a more stable patient representation and this would
reflect in their overall use of the CHI tracing service. There was also significant variation in the
dispensing rates (prescriptions items per day) of the five pharmacies.

The purpose of data validation was to ensure that the data deposited by any pharmacy contained, as a
subset, an accurate representation of the actual prescriptions dispensed and submitted to PPD for
payment and entered by MEMO in the normal course of data entry. Standard operating procedures in
MEMO did not permit the entry of:

   Stock orders, GP10A
   Dental prescriptions
   Hospital prescriptions
   FP10
   Drug problem centre prescriptions

In addition, private prescriptions and OTC dispensing were not notified to PPD or MEMO but were
sometimes entered onto the computer.

The data fields, described under section 2.3 MEMO services, were parsed automatically to yield a
prescription item record similar to that created by MEMO data entry staff:

Field                                Pharmacy        MEMO
Type (GP10, dental, etc.)            √               x
CHI number (or dummy)                √               √
Prescribing GP code (or dummy)       √               √
Pharmacy code (or dummy)             √               x
Prescribing date and time            X               √
Dispensing date and time             √               x
Prescribed medication                √               √
Prescribed amount                    X               x
Prescribed directions                X               x
Dispensed medication                 √               √
Dispensed amount                     √               √
Dispensed directions                 √               √
Exemption status                     √               x

Although the MEMO system was capable of collecting fields such as pharmacy code and exemption
status, the standard operating procedures (SOP) did not permit this so as to maintain entry throughput.
In addition, data entry is approximately 2 years in arrears and a suitable CHI database is chosen for this
entry. These SOP were not altered for this project. Under these SOP the pharmacy data was superior
particularly in regard to patient identification and date and time the drug was used by the patient.




                                                   26
We wished to establish that no software faults or procedures in the pharmacy shop had affected the
accuracy of the data. Although over 60,000 prescription records were available for automatic
comparison (at the time of performing the comparison) it was decided to perform an initial comparison
of 14,000 items by semi-automatic methods. This allowed closer inspection of the data. In particular,
for one pharmacy over a four-month period, a total of 74 deposits yielded 16,458 records consisting of
1,024 OTC, 37 private and 15,397 NHS. PPD bundle returns entered by MEMO gave 14,048 records
for the same period giving a discrepancy of 10% to be accounted for.

Pharmacy data and MEMO data without a defined CHI number were excluded from further
comparison. Those items indicated as unidentifiable from the prescription, not available on the
database (such as recently born babies) and temporary residents accounted for 610 pharmacy items and
614 MEMO items. This was a much higher rate of failure to allocate a CHI number to a patient than is
typical, but can be attributed to the high number of tourists (temporary residents) making use of the
chosen pharmacy (75% of the unidentified CHI numbers were allocated a temporary resident dummy
CHI).

The remaining records of these two data sets were merged in the following priority order:

CHI number (possibly dummy)
Preparation id - a cumulative hash of the components of the preparation description
Date and time
Record flavour - 'D' and 'E' for pharmacy data, 'M' for MEMO entered data
Amount
GP code (possibly dummy)

A cumulative hash offered a simple but effective method for equating preparation descriptions such as
"Losec 20mg Capsules" and "Losec Capsules 20mg" since addition is commutative. This merge proved
very effective in bringing pharmacy data items and MEMO data items together.

Patients with a one-to-one match in their item data accounted for 12,411 records of the pharmacy data
and 12,367 records of the MEMO data. The difference is accounted for by the dispensing of some
items as separate aliquots yielding extra pharmacy records. The total amount of medication matched
exactly.

The remainder of the 10% record count discrepancy was accounted for as follows:

(1) MEMO SOP did not permit the entry of certain types of prescription

Hospital prescription items (6 pharmacy records)
Dental items (46)
GP10A stock orders (335)
FP10s (7)

(2) Pharmacy data contained two or more records for items dispensed piecemeal (owings) whereas
MEMO recorded one (15 additional records)

(3) Pharmacy data contained redundant record pairs for corrected PMR items (927 records). (See
section on MEMO Services). There was a small excess of 'E' records (3) as a result of a rare software
fault in the JRC software.

(4) A single prescription can be dispensed in instalments over a period of time. 60 such prescription
items recorded by MEMO matched 205 records from the pharmacy.

(5) 10 OTC items were incorrectly recorded as NHS items when given a label by the pharmacy.

(6) "Edge effects" were anticipated with the windowed data. Consequently, pharmacists and their staff
were asked to synchronise prescription deposits with the beginning of a new PPD bundle.

One must remember that MEMO recorded prescribed date and time whereas the pharmacy systems
recorded dispensed date and time. If the comparison window is to contain prescriptions that can be



                                                   27
matched we must define the comparison window without reference to the date and time on the data.
Consider also that a small (2%) but significant fraction of the prescriptions arrive after the medication
has been dispensed and the PMR data recorded. Legally, this can be up to 24 hours after the pharmacy
receives the prescribing request by telephone or facsimile. These physical prescriptions may then find
their way into the next month's bundle since a PPD bundle must be submitted promptly at a designated
date. Thus an electronic record could exist for which no physical prescription was available. Similarly a
physical prescription could be submitted for which there was no electronic record available since the
deposit time was used for window membership rather than the time and date on the record (see above).

There is an issue here, since it is a subjective judgement as to whether a record is unmatched through
error or through the edge effect. A total of 143 pharmacy records and 321 MEMO records were
unmatched and judged to result from the "edge effect".

(7) The physical prescriptions received by MEMO were available for resolution and classification of
discrepancies. Preparation and amount errors occurred in 14 pharmacy records and 72 MEMO records.
In addition, pharmacy staff often found it easier to recreate a label afresh rather than edit the previously
recorded PMR entry. This double entry accounted for 85 pharmacy records.

(8) There were records from the pharmacy (583) and MEMO (614) that were (consistently) identified
but did not have a match. This was the result of mistaken allocation of CHI numbers within the
pharmacy and/or MEMO. MEMO SOP prevented the use of an up-to-date CHI database for data entry
within MEMO and it was likely most of the error arose here. A sample of these data were carefully
checked and confirmed the suggestion that the pharmacy records were substantially more likely to be
correct. An alternative system was available in MEMO to enter the pharmacy's data with the
appropriate CHI database 'snapshot' but this would have meant re-entry of this data on the normal
system at a later date. Cost implications ruled this out.




                                                    28
                               Conclusions
The present project has demonstrated that within community pharmacies the capture
of electronic prescription data with accurate CHI number allocation:

   Was technically feasible
   Worked well in practice
   Did not impact significantly on pharmacist time to dispense a prescription.
   Was viewed positively by pharmacists & pharmacy technicians
   Uncovered some issues that can be overcome by technical or procedural
    modifications
   Provided accurate on-line patient-specific dispensed prescribing data


The success of the present project has major implications for the NHS in Scotland.
Implementation of a similar system throughout Scotland would reap significant
rewards in:

 Improved patient compliance
 Improved pharmaceutical care
 Improved drug interaction checking
 Improved prescribing audit and formulary control
 Improved drug utilisation data / trends
 Improved fraud detection
 Improved capture of dispensing data by the Pharmacy Practice Division and
  reduced costs of this
 Improved pharmacovigilance through record-linkage
 Improved outcomes research
 Improved pharmacoeconomic evaluations of new pharmaceuticals




                                         29
Project publications and Presentations to date
McGilchrist MM, MacDonald TM. Automatic Capture of community prescribing. International Society
of Pharmacoepidemiology, Amsterdam, August 1996. Pharmacoepidemiology and Drug Safety 1996;5
(Suppl 2):S37.


McGilchrist MM, MacDonald TM. Automatic capture of community prescribing. International Society
of Pharmacoepidemiology, Orlando, Florida August 1997. Pharmacoepidemiology and Drug Safety
1997;6 (Suppl 2):S78


McGilchrist MM, MacDonald TM. Quality control analysis of prescription data captured automatically
from community pharmacies. International Society of Pharmacoepidemiology, Berlin, August 1998.
Pharmacoepidemiology & Drug Safety 1998;7 (Suppl 2): S190, Abstract 262.


McGilchrist MM, Morris AD, Davey PG, McDevitt DG, MacDonald TM. Patient-specific PACT data:
available now. BMJ;1998:1530-1.




                                                30

								
To top