Docstoc

Ghana DCE May 13

Document Sample
Ghana DCE May 13 Powered By Docstoc
					    Rural practice preferences of
    medical students in Ghana: a
     discrete choice experiment


           Margaret E. Kruk, MD, MPH
   Department of Health Policy and Management
Columbia University Mailman School of Public Health
Research team
                      Margaret E. Kruk
                     Jennifer C. Johnson
                      Mawuli Gyakobo
                     Peter Agyei-Baffour
                        Kwesi Asabir
                        S. Rani Kotha
                       Janet Kwansah
                      Emmanuel Nakua
                       Rachel C. Snow
                    Mawuli Dzodzomenyo

Kruk ME, Johnson JC, Gyakobo M, et al. Rural practice preferences
among medical students in Ghana: a discrete choice experiment. Bull
World Health Organ. May 2010;88(5):333-341.

                                                                      1
Agenda
 Discrete choice experiments
 Human resources for health in Ghana
 Methods
 Results
 Discussion




                                        2
Agenda
 Discrete choice experiments
 Human resources for health in Ghana
 Methods
 Results
 Discussion




                                        3
Establishing preferences: random
utility theory
   Random utility theory states that:
    ◦ A good or service can be described on the basis of its
      characteristics
    ◦ The individual’s value for the good or service depends
      on the levels of those characteristics
 Assumes that users are economically
  rational and utility maximizing
 Utility is a latent construct; all we observe
  is the choice made


                                                           4
  Random utility model
      Utility of an alternative (of several options) is a
       function of its component attributes and
       attribute levels such that Yiq is the utility of
       individual q for the ith alternative:

                                    Yiq  Xi i  iq

       where Xiq is the vector of attributes for the ith
       choice facing individual q and εiq is the error due
       to taste heterogeneity and measurement error

Lancaster K. A new approach to consumer theory. The Journal of Political Economy. 1966;74:132-157
                                                                                                    5
               Utility function
            Dependent variable is alternative chosen (A, B,
             or C)
            A generic utility function is modeled as:

                 U  11  2 2  3 3  ... n 1
               where U is the change in utility from moving
               from alternative A to alternative B, X is the
               difference in attribute levels between A and B,
               and  are the coefficients to be estimated.


Breidert, Hahsler, & Reutterer, 2006
                                                                 6
Analysis
   DCE data can be modeled using:
    ◦ Multinomial logit
    ◦ Conditional logit (conditionality within tasks)
    ◦ Mixed logit or probit (accounts for correlated
      observations within respondents and conditionality
      within tasks)
    ◦ Hierarchical Bayesian models (as above and estimates
      preferences at individual level)
   Results are utility coefficients or part-worths
    (interval data)



                                                             7
Revealed versus stated preferences
 Can assess utility using revealed or stated
  preferences (or combination)
 Revealed data demonstrate actual behavior
  and so are preferred by
 But revealed preference data have
  limitations, particularly in health care:
    ◦ Markets are imperfect or non-existent; subsidies
      mask true costs
    ◦ Regulations limit range of behavior
    ◦ RP data difficult to disaggregate into single
      attributes
    ◦ RP data cannot incorporate unavailable attributes

                                                          8
Discrete choice experiments
   One of several methods of assessing stated
    preference
   Akin to other contingent valuation methods
    such as standard gamble
   High internal consistency and test-retest
    reliability
   Results consistent with a priori expectations
    (criterion validity)
   Also consistent with results of related
    instruments such as standard gamble
    (convergent validity)

                                                    9
Discrete choice experiments
   Can obtain willingness to pay by calculating
    marginal rate of substitution between an
    attribute and the price/cost variable
   May be more realistic than simple rankings or
    ratings as it incorporates multiple attributes
    thereby better approximating real life consumer
    decision making
   Easy to administer (choose to “buy” A or B)
   Used in marketing, transport, environmental
    economics
   Increasingly used in health care

                                                  10
              DCEs in health have been used to:
                Elicit physician preferences in priority setting
                Obtain patient preferences for structure of
                 physician practice
                Identify preferred clinical management
                 strategies
                To indicate relative value of health services
                 to patients
                   ◦ Unlike C/E analysis takes account of relative
                     importance of different aspects of health care:
                     structure, process and health and non-health
                     outcomes

Ryan M, Farrar S. Using conjoint analysis to elicit preferences for health care. Bmj. Jun 3 2000;320(7248):1530-153311
DCEs in HRH
 Chomitz K, Setiadi G, Azwar A, Ismail N, Widiyarti. What do
  doctors want? Developing incentives for doctors in serve in
  Indonesia's rural and remote areas. Washington, DC: World
  Bank; 1998.
 Hanson K, Jack W. Health worker preferences for job attributes
  in Ethiopia: Results from a discrete choice experiment. Washington:
  Working paper, Georgetown University; 2008
 Kolstad JR. How to make rural jobs more attractive to health
  workers. Findings from a discrete choice experiment in
  Tanzania. Health Econ. Jan 21 2010;21:21.
 Blaauw D, Erasmus E, Pagaiya N, et al. Policy interventions
  that attract nurses to rural areas: a multicountry discrete
  choice experiment. Bull World Health Organ. May
  2010;88(5):350-356.

                                                                    12
Agenda
 Discrete choice experiments
 Human resources for health in
  Ghana
 Methods
 Results
 Discussion




                                  13
Ghana




        14
Ghana
 Population 22.2 million; 62% in rural areas
 GNI $1320 PPPs per capita
 Major urban rural disparities in
  infrastructure and health service
  utilization
 80% of urban households have electricity;
  31% of rural do
 85% of urban women deliver in a health
  facility; 39% of rural women
                                                15
Human resources for health and
outmigration
   2442 MDs were working in Ghana in 2009
   Huge source of emigrant physicians: 61% of
    medical school graduates between 1985 and
    1994 emigrated, primarily to UK and US
   This has slowed recently, attributed to
    increase in salaries
   Additional Duty Hours Allowance increased
    salaries by 75-150% to approximately
    $14,000 annually
                                                 16
Geographic distribution
 69% of physicians practice in Accra region
  or the Kumasi teaching hospital (Komfo
  Anokye)
 Physician to population ratios:
    ◦ 1:5000 in Greater Accra region
    ◦ 1:92,000 in Northern region




                                               17
Medical education in Ghana
   Medical education in Ghana consists of:
    ◦ 3 years of basic sciences
    ◦ 3 years of medical studies
    ◦ 2 year housemanship in which students rotate through
      general medicine, obstetrics and gynaecology, surgery and
      paediatrics
 There are 4 medical schools in Ghana: the University of
  Ghana (UG), Kwame Nakrumah University of Science
  and Technology (KNUST), University for Development
  Studies (UDS) and University of Cape Coast (UCC).
 The UCC medical school began accepting students in
  2007 and had no fourth year students yet.
 At the time of the study, all fourth year students in the
  country were training at UG or KNUST

                                                                  18
Postings
 After completing housemanship, medical
  students are provided with available MoH
  postings; majority of these are in under-
  served areas (rural or peri-urban)
 There is no return-of-service obligation
  (bonding) for vast majority of students




                                              19
Research question
   What job attributes influence senior
    medical students’ choice of rural practice
    posts?




                                                 20
Agenda
 Discrete choice experiments
 Human resources for health in Ghana
 Methods
 Results
 Discussion




                                        21
 Designing a DCE
1.   Identify characteristics (attributes)
     ◦   Policy options, focus groups, literature
2.   Assign levels to attributes
     ◦   Plausible, can be cardinal, ordinal, categorical
3.   Choose subset of scenarios
     ◦   Experimental design used to reduce number
         while maximizing efficiency
4. Establish preferences
5. Analyze data


                                                            22
Attributes
 Identified long list of policy-ameable
  attributes from extant literature
 Conducted 7 focus groups with 3rd and
  5th year medical students at UG and
  KNUST
 Discussion on career plans, motivation for
  rural practice, important attributes to
  encourage rural practice, trial rankings of
  attributes
                                            23
Attributes and levels




                        24
Selecting a subset of scenarios
 Total of 384 possible alternatives
 Assumed independence of attributes
 Selected scenarios that maximized
  orthogonality (low correlation between
  levels of attributes), maximized level
  balance, and minimized overlap among
  levels within one task (efficient design)
 Selected 22 alternatives (paired for 11
  tasks) and 1 fixed task using Sawtooth
  Software
                                              25
26
Fielding
 Invited all 4th year medical students in
  Ghana to participate
 Gave electronic survey on background,
  career plans, motivation for rural practice
  along with DCE module (12 choice tasks)
  in computer labs with trained surveyors




                                                27
Agenda
 Discrete choice experiments
 Human resources for health in Ghana
 Methods
 Results
 Discussion




                                        28
Response
 Out of 310 fourth-year students enrolled
  in Ghana’s medical schools,
 307 (99.0%) students participated in the
  survey
 5 survey files were corrupted by viruses
  or lost due to computer malfunction
 Analysis conducted with 302 total
  records
 The survey took a mean of 31.6 (SD
  12.45) minutes.
                                             29
Demographics




               30
Demographics




               31
International and rural practice




                                   32
Mixed logit results Model 1




                              33
Mixed logit results Model 2




                              34
Mixed logit results Model 2




                              35
Policy simulations




                     36
Agenda
 Discrete choice experiments
 Human resources for health in Ghana
 Methods
 Results
 Discussion




                                        37
Summary of findings
 Students valued rural job attributes that
  enabled them to perform well clinically
  (improved infrastructure and equipment)
  and enabled their professional growth
  (supportive management)
 This was equivalent to a bonus of 100% of
  base starting salary
 Consistent with focus group findings


                                          38
Supportive management
 Supportive work culture and management
  especially important to women—one of
  only two attributes for which there was a
  gender difference
 Consistent with results of some studies
  showing supportive management
  increased motivation



                                          39
Two year contract
 Highly valued: this was echoed in focus
  groups
 Possible that some students would be
  willing to accept a guaranteed short-term
  placement (with other incentives)




                                              40
Housing
 Basic housing is considered a pre-
  requisite for rural practice by students
 This is due to low availability of quality
  housing in rural towns in Ghana and
  students’ awareness that this is a standard
  offering by rural hospitals
 Consistent with Hanson and Jack study in
  Ethiopia


                                            41
Salary
 High utility, particularly for bonuses of 50-
  100% salary
 Literature mixed on salaries needed to
  change behavior (may be more important
  to practicing MDs than medical students)
 Students may also be willing to forego
  urban salary for a brief rural experience,
  particularly if well supported


                                              42
Other factors
 Allowances for children’s education not as
  important; possibly as students were
  young
 Car not important—relatively well off
  students may have or expect to buy their
  own car




                                           43
Research to policy?




                      44
45
                                                         Relative value of incentives, compared
                          2.5                             2.4   to a 50% salary increase

                          2.0                                                       2.0
                                                                                                         1.7

                          1.5
                                                                                                                             1.3
                                                                                                                                                  1.2
                                                                                                                                                                       1.1
                                          1.0
                          1.0



                          0.5



                          0.0
                            50% salary increase                             Supportive                                  Utility car                         Superior housing
                                                                            management



Reference: Kruk M, Johnson J, Gyakobo M, Agyei-Baffour P, Asabir K, Kotha R, Kwansah J, Nakua E, Snow R, Dzodzomenyo M. Preferences for rural practice incentives among medical
students in Ghana: A discrete choice experiment. Bull WHO. Submitted 1 Oct 2009.                                                                                                  46
Validation?
   Do a policy experiment that includes top
    incentives and compares against standard
    offering but need to keep in mind…




                                               47
Validation?
 Policy resistance (2 year contract!)
 Civil service regulations
 How to measure effect with small n/low
  variation?




                                           48
Other uses of DCE….




                      49
Preferences for delivery facilities in
Tanzania
   Att ibute
      r                                                Ut a,b
                                                        ility   p-value
   Distance
    1/2 hour by foot                                    12.4    <0.001
    1 hour by foot                                      11.9    <0.001
    1 and 1/2 hoursby foot                              12.8    <0.001
    2 hours by foot                                     2.8      0.018
    3 hours by foot                                      0        ref
   Type of provider
    Do  ctor                                            29.0    <0.001
    Clinical officer                                    6.9     <0.001
    Nurse                                                0       ref
   Provider   attitude
              r             s
    Provide smiles, listen carefully                   168.5    <0.001
              r
    Provide doesnot smile, does not listen carefully     0       ref
   Ava  ilabilityof drugsand med         ment
                                ical equip
                               nt w
    Drugsand medical equipme al ays av     ailable     160.0    <0.001
                               nt         ys
    Drugsand medical equipme not alwa available          0       ref
   Ava  ilabilityof tra nsport
    Transport av    ailable                             21.5    <0.001
                  not
    Transport available                                  0       ref
   Cost
    250 Sh   illings                                    33.3    <0.001
    500 Sh   illings                                    46.7    <0.001
    1000 Sh    illings                                  10.8    <0.001
    2000 Sh    illings                                  20.2    <0.001
    3000 Sh    illings                                   0       ref

                                                                          50
          Policy simulations
    ar
     io
   Scen                                                             s
                                                            Home Dis penHe C t
                                                                     ary lth er
                                                                         a en                              t
                                                                                                           al
                                                                                                         Hos pi
                                                            % (SD)% (SD)% (SD)% (SD)
             b
    B l
      s ine
   1.ae                                                      ( )
                                                             2.5
                                                           55.8            3.3 )
                                                                             (
                                                                             0.7          22.0
                                                                                           (
                                                                                         28. )             ( )
                                                                                                           1.3
                                                                                                         12.7
       e fre
          er
           tans
   2. Provid port
       h
       ea
      To lthn
            ter
             ce                                              ( )
                                                             2.5
                                                           53.8            2.8 )
                                                                             (
                                                                             0.6          22.2
                                                                                           (
                                                                                         36. )            2
                                                                                                         7.(0.8)
       hosa
      To itlp                                                ( )
                                                             2.5
                                                           53.2            2. 0.
                                                                            9 7)
                                                                             (             (1.7) 22.5
                                                                                         21.4      1.8
                                                                                                   ( )
    R c ad
       e lel    t
                s
   3.edu totivery cos
      Co hlt c t 250 c
       t e
       sofah ener: TZS                                       ( )
                                                             2.5
                                                           53.3            3. 0.7
                                                                            2 )
                                                                             (            12.
                                                                                           (
                                                                                         33. 2)            ( 2)
                                                                                                           1.
                                                                                                         10.4
       sof pa
       t
      Co hos:250
           il
            t  TZS                                           ( )
                                                             2.5
                                                           53.8            3.1 )
                                                                             (
                                                                             0.7          41.
                                                                                           (
                                                                                         19. 5)            ( 7)
                                                                                                           1.
                                                                                                         23.7
       e drugs
           and en d s i
             quip
   4. Provid e mt in   ar
                        es
                    ispen                                  50. 6)
                                                            62.
                                                             (             82.2
                                                                            (
                                                                          38. )           6.7 )
                                                                                            (
                                                                                            0.7            ( 6)
                                                                                                           0.
                                                                                                         3.9
           ra t providein
            tti p
   5. Improveude/    c d
                      e i
                 erforman re a 27. )
                        spens
                           is    (
                                91.8                                        (2.0) 26.
                                                                          33.9     4        ( 3)
                                                                                            1.
                                                                                    (2.0) 11.8
        e drugs mt
            and en          provi
   6. Provid equipand improve der
   t tud
    t
   ai e/performa s i
             n
             cedin ar
                ispenes            9. 1.
                                    9 4)
                                     (                                      ( )
                                                                            1.7
                                                                          82.0            5.
                                                                                           8(1.0)          ( 5)
                                                                                                           0.
                                                                                                         2.3
        e drugs mt
            and en          provi
   7. Provid equipand improve der
                ie
         endis r a elthn s andt ls
           i pens ,
   performanc sha       ce,    ia
                        ter hos p    ( 3)
                                   9.41.                                  65.3
                                                                            ( 5)
                                                                            1.           14.4
                                                                                           0.8
                                                                                           ( )           10. )
                                                                                                           (
                                                                                                          90.8




Kruk ME, Paczkowski M, Mbaruku G, de Pinho H, Galea S. Women’s preferences for place of delivery in rural
Tanzania. a population-based discrete choice experiment. American Journal of Public Health. 2009. 99(9): 1666-72.
                                                                                                                    51
           External validity
                         al eof iv
                          l a lt
                           pc ad
                      Actu sel ery Predi rd
                                       c l e futu ery
                                       te c
                                        d ofe iv
                                          p
                                          a     el
                                        ted re c
                                         a ey ts cho
                                         t
                                     Surv Disce   e
                                        ce
                                         forexperi
                                     preferen t men
             Pl of a DHS 1999 t studyery u g
              e li
              c ev
             a d ery   c ,e   d,e
                          DHS 2004 e f d sim
                                 Curren li
                                     fut v
                                      ure      a
                                               ltion
              (b
              %)
             Home                    74.966.
                                          4                         5
                                                                   62.               8.3               55.8
              s (%)
               ary      4
             Dispen15.716.                                         1
                                                                   59
                                                                    .               41.8               3.3
               a n(
               H ter
               eth ce
                l  %) 3.1                           6.4            1
                                                                   13
                                                                    .                7
                                                                                    29.                 2
                                                                                                       28.
              t (%) 6.
              al
             Hospi  3                              10.
                                                    9               3
                                                                   10.              20.2               12.7




         DCE was a more accurate “predictor” of place of
          most recent delivery than direct question about
          place of future delivery in the survey

Kruk ME, Paczkowski M, Mbaruku G, de Pinho H, Galea S. Women’s preferences for place of delivery in rural
                                                                                                  99(9): 1666-72.
Tanzania. a population-based discrete choice experiment. American Journal of Public Health. 2009.52
Clinic preferences for episodic illness in Liberia
   Attribute                                Estimated utility (SE)
   Waiting time (in hrs.)                            -0.002 (0.020)
   Clinic workers respectful                          0.488 (0.066) ***
   Medicines always available                         2.135 (0.128) ***
   Good physical examination                          4.398 (0.217) ***
   Cost (x1000 Liberian dollars1)                    -0.326 (0.083) ***
   Government management                              0.093 (0.055) *


   Number of respondents                                     1,414
   Number of observations                                   21,894
   Log likelihood                                        -3,739.63
   Likelihood ratio χ2                                    1,511.18

*p < 0.10, **p < 0.05, ***p < 0.01
11 US dollar = 62 Liberian dollars (2008)
                                                                          53

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:4
posted:7/1/2012
language:English
pages:54