Received: 5 June 2018 Revised: 28 August 2018 Accepted: 29 August 2018

DOI: 10.1002/hpm.2671

RESEARCH ARTICLE

Policy intervention for access to medicine: Does it

work similarly for poor and non‐poor?
Arijita Dutta1 | Satarupa Bandyopadhyay2

1
Department of Economics, University of
Calcutta, Kolkata, India Summary
2
Bethune College, University of Calcutta, Many federal states in India have recently taken steps to
Kolkata, India
improve distribution and affordability of medicines across
Correspondence
Dr Satarupa Bandyopadhyay, Bethune College,
public hospitals for reducing high out‐of‐pocket (OOP)
University of Calcutta, 181, Bidhan Sarani, expenditure. West Bengal has introduced a Public‐Private
Kolkata 700006, India.
Email: satarupabandyopadhyay17@gmail.com
Partnership (PPP) scheme, Fair Price Medicine Shops (FPMS)
Funding information within government hospitals in 2012, which offers high dis-
National Rural Health Mission and Depart- counts on maximum retail price of the drugs. This model
ment of Health and Family Welfare, Govern-
ment of West Bengal, India introduces the state as a facilitator, rather than provider or
financer, of health care. This paper attempts to measure its
impact on OOP expenditure (OOPE) of patients using pro-
pensity‐score‐matching technique on the data collected from
primary survey among patients. The study finds that although
for non‐poor patients, the average OOPE has reduced signif-
icantly, the impact has been counter‐productive for the poor
patients, hinting that PPP intervention seems to work only
for relatively better‐off people, as the best alternative for
the poorest remains to be free provision of drugs from the
government. The difference in outcome lay in the fact that
the nature of control groups differed between poor and
non‐poor patients. Those poor patients who did not visit
FPMS received most of the medicines free (hence OOPE
nearly zero), while for the non‐poor patients not visiting
FPMS bought the drugs from outside retail shops.

KEYWORDS

access to medicines, poor and non‐poor, propensity score matching,

public private partnership, West Bengal

The work was carried out in the Department of Economics, University of Calcutta, India.

Int J Health Plann Mgmt. 2019;34:e557–e568. wileyonlinelibrary.com/journal/hpm © 2018 John Wiley & Sons, Ltd. e557
e558 DUTTA

1 | I N T RO DU CT I O N

Access to medicine is defined as having medicines continuously available and affordable at public or private health facilities
or medicine outlets that are within one hour's walk from the homes of the population.1 According to WHO estimates, nearly
a third of the world's population lacks access to the most basic essential medicines, while in the poorest parts of Africa
and Asia this figure climbs to a half.2 Additionally, in developing countries, pharmaceutical drugs now account for 30%
to 50% of total health care expenditure, compared with less than 15% in established market economies.3 Although the
access to essential medicines worldwide increased from covering roughly 2.4 billion to 4.3 billion between 1975 and
1999,4 a closer look reveals that the overall number of those without access remained almost unchanged and that these
people are primarily the poorest and the most marginalized. Limited access to professional health services and aggres-
sive marketing of drugs on an unregulated market have increased wastage of scarce financial resources and hence
incidence of poverty, particularly among those who are already poor.5 Eventually, this trend of lack of access and high
out‐of‐pocket expenditure (OOPE) on drugs leave the poor either to forego treatment (hence losing man days, produc-
tivity, and income), or to seek the same in private and/or unregulated drug market often leading to catastrophic medical
expenditure (when medical expenditure for one illness episode is more than 40% of Annual Non‐food expenditure of
the family), pushing them to medical poverty trap. The issue of user charges for public services has opened up a greater
debate since the introduction of its concept in World Bank.6 While this was a common strategy for many developing
countries entrenched into health policy reform at the background of macroeconomic structural adjustment, this has
been thoroughly criticized for being extremely regressive in nature lacking pooling of risks. Although it is expected
to offer cross subsidy from richer to poorer patients, often the impact of user charges on the poor has been insufficient
to provide financial protection and hence it turned out to be regressive in many countries.7,8 The result might get
worse, in the presence of high transaction costs particularly among the poor.9
Geographically, the lack of access to essential medicines is especially severe and concentrated in Africa and India.
In fact, 38% of the people without access to essential medicines live in India, and 15% live in Africa.4 India, known as
the “global pharmacy of South,”10 has 65% population suffering from lack of access to medicine, primarily due to
absence of sound public distribution facilities.11 Studies identify that this deficiency in access to medicines has
resulted in the highest share of OOPE on health care, arguably the most regressive kind of health finance option.12
Using Consumer Expenditure Survey data from the National Sample Survey in India, Garg and Karan13 estimates that
purchase of drugs constituted 70% of the total OOPE on health, and approximately 32.5 million persons slipped
below the poverty line in 1999 to 2000 due to this. The study by Dutta et al in 201414 revealed that in 2011 only
6% in in‐patient department (IPD) and 10% in out‐patient department patients in secondary public hospitals in West
Bengal received all the medicines free of cost from the hospital, primarily owing to the demand‐supply mismatch and
failure to estimate demand correctly.
Although lack of access to essential medicines had been a common problem faced by all Indian states, very few
of them came up with specific policy to ensure easy availability and affordability of drugs, other than the over‐arching
framework of National Rural Health Mission for providing universal access to health care since 2004. Delhi (1992)
and Tamil Nadu (1994) were the first two states introducing state‐specific policies for ensuring better access to med-
icines through more leak‐proof and strong monitoring systems. In the Tamil Nadu model, considered to be the bench-
mark model to provide free medicines to every patient visiting public health facilities, autonomous public sector body
named Tamil Nadu Medical Services Corporation procures high quality generic medicines through a transparent bid-
ding system, which is then supplied to public health facilities through a demand sensitive coordination. Bihar, Delhi,
Madhya Pradesh, Kerala, and Rajasthan also came up with similar, although differentiated, models.15 In 2011, Rajas-
than followed the Tamil Nadu model by establishing Rajasthan Medical Services Corporation and introducing the
Mukhyamantri Free Fund Scheme, and within 2 years, an overall improvement in access to health care, financial risk
protection, and health system expansion was observed.16 All these models succeeded to improve access to medicines
at different degrees, although they required huge public investment in the gamut, which could be possible to under-
take only in states with relatively healthy budgetary positions or with generous central grants.
DUTTA e559

With nearly one quarter of her population lying below poverty line in 2004 to 2005,17 people in the state of
West Bengal depend quite heavily on the public provided health care facilities. The state more than doubled its
per capita drug expenditure within a decade (from 2000‐2001 to 2010‐2011),11 although highly skewed distribution
of funds offset the expected gain from the enhanced investment. The aforementioned report of WHO15 shows that
in West Bengal, 70% of drug expenditures were spent in urban and/or tertiary care, representing a clear organiza-
tional failure in distribution. As a result, people in West Bengal had to spend significantly as OOPE in buying medi-
cines as most of the benefits from public subsidy in medicine was grabbed by the relatively richer section of the
population.18However, with existing huge public debt and hence extremely narrow fiscal space, heavy investment
following Tamil Nadu Medical Services Corporation model ensuring free access to drugs in this over populous state
did not seem affordable.
In this backdrop, the government of West Bengal ushered in an entirely different model in the health care sector
in 2012 by launching Fair Price Medicine Shops (FPMS) across secondary and tertiary hospitals in the state under
Public‐Private Partnership (PPP) framework.19 In tune with recommendations by Dutta and Bandyopadhyay,20 the
program ideally aimed to combine two potentially efficient actors for the procurement and distribution of medicines,
namely public and private, to develop a synergic system to create access to affordable and high quality medicines to
patients. It is basically an output‐based contracting‐out model, where the contracts are offered to private pharmaceu-
tical retailers by online bidding depending on the percentage of discount for each hospital. The bidder with the
highest discount gets the contract and is expected to give the stipulated rate of discount on maximum retail price
over all the items across the board for all patients, either visiting the public hospital or any outside institution with
a valid prescription. In order to have a check on the minimum quality, the state authority came up with a list of phar-
maceutical producers with specific turnover history, from where the FPMS can procure medicines. The uniqueness in
this model is that:

a. The role of the state is a regulator and not provider or financer as in all other states
b. There is zero public investment required, rather state earns revenues as rents from the FPMS
c. Beneficiaries are not only the patients going to the public hospitals (where the FPMS are located), but anyone
having a valid prescription.

The large discounts offered on MRP by FPMS are a viable business strategy because it precisely passes on a part
of the markups that it itself receives directly from the producer, thus bypassing a few intermediaries in the distribu-
tion chains. The reduced price, in no way, is a signal of lower quality, as campaigned vocally by a group with economic
and/or political vested interests; rather, it stands for efficient distribution practice involving public and private agen-
cies. On the other hand, FPMS does not guarantee to offer each generic at a price lower than all available branded
alternatives of that generic. They aim to offer drugs at a reduced rate compared with the branded market leaders
(fastest moving brands in that particular drug segment).
In the First Phase in 2012, 35 FPMS were established with discounts ranging from 48% to 67.25% on MRP, and
by end of 2014, 94 FPMS (out of 116 planned) became functional. The FPMS were directed to stock 142 mandatory
medicines. These medicines were allowed in the form of branded generic which is a new term coined in Indian phar-
maceutical market, which essentially carries the brand names of the companies, but does not go through any specific
marketing or promotional activities and expenditures. They are directly offered to retailers and hospitals, bypassing
the longer chain of other intermediaries. Many often, this category of drugs is marketed in same name by a company
as the usual branded category. These medicines are supplied in bulk primarily to hospitals and retail chains. An initial
evaluation by Dutta and Bandyopadhyay21 revealed that on the totality the program was successful in reducing
OOPE among patients seeking treatment in public hospitals. However, the aim of this paper is to evaluate the possible
differentiated impact of implementation of this PPP model on the poor and non‐poor patients admitted in a few state‐run
hospitals in West Bengal and thus to locate if this kind of low‐cost policy for the government can work to support the most
vulnerable groups to curb the possibility of medical poverty trap.
e560 DUTTA

2 | MATERIALS AND METHODS

With the abovementioned objectives, the study utilizes primary data composed of patients' exit interviews collected
through well‐structured questionnaires for 1000 hospitalized patients at the IPD. Crucially, for each of these patients,
the prescriptions were copied. To capture the total OOPE of hospitalization, the inpatients were interviewed just
before their departure from the hospital after receiving the discharge certificate, so that data for the entire illness
episode can be captured.
Out of the 94 hospitals, 24 were identified, where FPMSs were launched in the first phase. These hospitals were
divided into three categories: good, bad, and medium performing in terms of total sales volume of FPMS in the month
preceding the survey. Three hospitals were selected from each category (SSKM Medical College, Barasat District
Hospital, and Balurghat District Hospital from good category; Chittaranjan Medical College, Bankura Sammelani
Medical College, Krishna Nagar District hospital, and Baruipur sub divisional hospital from medium category; and
Bardhaman Medical College and Jalpaiguri District Hospital from bad category) randomly. In total, nine hospitals were
selected, with four of them being teaching medical colleges, five being district hospitals, and one being sub‐divisional
hospital. Four departments were selected, namely, general medicine, cardiology, orthopedic, and pediatrics from each
hospital, primarily keeping in mind the varied drug intensity of treatment in different departments, captured during
initial Focus Group Discussions with doctors. Since district and sub‐divisional hospitals do not have cardiology
departments, eye departments were considered in those cases. Equal number of patients were surveyed in each of
these hospitals. To identify the differential impact among poor and non‐poor patients, patients are subdivided in
two groups based on Monthly Per Capita Expenditure. The survey collected data on total family expenditure per
month and number of family members, with which the Monthly Per Capita Expenditure was calculated. This figure
was then categorized into poor or non‐poor categories using official poverty lines defined in Planning Commission
Report22 in rural and urban areas separately. Out of 1000 patients surveyed, 548 belonged to the poor category,
whereas the rest 442 were non‐poor. The final survey was conducted during November 2013 to August 2014.
Since the study design was explicitly quasi‐experimental, our sample has two analysis groups: one treatment
group and one control group. The treatment group consists of patients who had visited the FPMS to buy the med-
icines, and the control group is made up of those who did not visit FPMS, even if they went to other retail shops out-
side the hospital premises to buy medicines without any discount on MRP. Both treatment and control groups come
from a group of 803 patients in IPD, who had the “need” of buying one or more medicine as they did not receive all
medicines prescribed free of cost from the hospital pharmacy. Out of these 803 patients, who comprise the potential
FPMS visitors, 446 are poor and the rest are non‐poor. The main hypothesis that we test in this paper is that the pro-
gram effect was successful to reduce the magnitude of OOPE for the treatment group compared with control group
for both poor and non‐poor patients equally. Simply comparing between the means of OOPE by treatment and con-
trol groups shows that the former's average expenditure is far higher (statistically significant) than the control group
in case of both economic groups of patients (Table 1). In order to correct for the possible sample bias of self‐
selection, we need to use certain methodology for culling out the effect of program.
We have used propensity score matching because of its effectiveness in simulating an experiment and measuring
treatment impacts in non‐experimental studies, especially only for studies observing the post‐treatment effects, like
this one.23 In short, we first calculate the propensity of each individual to receive the treatment in the sample and

TABLE 1 Average medicine cost (INR) by treatment and control groups

Overall Poor Non‐Poor

Average medicine cost for the hospitalization episode for the patients who went 3580 3133 4089
to FPMS to buy medicine (Treatment group) (in INR)
Average medicine cost for the hospitalization episode for the patients who did 1860 1392 2614
not go to FPMS to buy medicine (Control group) (in INR)

Source: Analysis of primary data

DUTTA e561

club one treated individual and one individual from control group with similar propensity to compare their outcome
variable. This way, each treated person is compared with a non‐treated person, and the average difference of the
outcome variable is taken as average treatment effect on the treated. The main feature of the technique is that it
is expected to remove the selection bias by balancing the sample on the characteristics that potentially determine
the possibility of selection of patients into the treatment group. Balancing for selection bias (here it is self‐selection
because FPMS was available for all the patients) is done using a probit model, whose estimates are used to obtain the
propensity scores for selection into the program24 (Table 2). Propensity scores are the predicted probabilities to esti-
mate the probability of each case receiving the treatment (ie, going to FPMS or not). Once the scores have been gen-
erated, we have checked if the selection bias has been removed by performing bi‐variate tests against the treatment
variable on all the variables used to remove selection. Since the differences for treatment indicator are insignificant
after balancing, we conclude that the selection bias had been successfully removed, and our subsequent models
would be balanced. We also assess whether the “standardized bias”—the differences in means between treated
and matched clone from the control group divided by the square root of the average of the sample variances of
the two groups—was less than 25%.25 We have used nearest neighbor matching technique. This is expected to cap-
ture the average treatment effect on the treated in terms of the outcome variable.
Following WHO 2011,26 the paper identifies the three‐dimensional cubic concept of any program: breadth or
coverage, depth of services, and height of financial protection. It attempts to locate the breadth (population covered)
and depth (services received) of the policy intervention for poor and non‐poor patients. For the final outcome vari-
able of the height of the program (measure of financial risk protection), data on OOPE on item‐wise estimates for
expenses incurred during hospitalization, drugs, health care products (eg, blood, saline etc), diagnostic and laboratory
tests, travel, food and lodging for accompanying person, attendant charges etc were collected in the survey for the
specific hospitalization episode. Costs related directly to treatment were included in the estimates of direct medical
costs to the patient.
Since some provider‐level characteristics are likely to influence the choice to visit FPMS, we consider three such
categorical variables here: specific hospital, specific department, and share of medicines prescribed in their generic
names in the full list of prescribed medicines. The former two factors are important as they play a crucial role on mak-
ing the patients aware about the presence of FPMS through notice boards and communication from doctors and
other hospital staffs. The last one actually broadens the chance of getting the medicine under any brand name avail-
able at FPMS, while medicines written in specific brand names restrict the chance (Although officially the trained
pharmacist at the FPMS enjoys the right to change the brand name, keeping the basic generic unchanged, in many
cases the doctors are reported to be refusing to accept another brand, given their high vested interest.).

TABLE 2 Breadth of coverage and depth of services received

Non‐
All Poor Poor

Total patients surveyed 1000 548 452

Average share of generics in the prescriptions % 52.38 50.27 54.47**
Patients who needed to buy medicine from outside 803 446 357
Patients who visited FPMS (% share out of those who needed to buy medicines from outside) 64.51 61.05 71.37***
Patients buying medicines from FPMS 23.93 24.25 23.33
i) all medicines prescribed 60.12 62.28 56.11
ii) some medicines prescribed 15.95 13.47 20.56
iii) no medicines prescribed
Average number of medicines bought from FPMS 0.61 0.91**

Poor–non‐poor difference significant at:

***1% level; **5% level.
Source: Analysis of primary data.
e562 DUTTA

At the patient level, nine variables were considered to capture their socio‐economic characteristics. They are as
follows: geographical location of the household (rural or urban), caste of the household (general or reserved), type of
fuel used in the household (green or red), type of sanitation facility in the household (improved or non‐improved),
patient's years of education, and regular access to newspaper of the patient (no or yes). While the initial ones repre-
sent the household's social, economic, and living conditions, the last two stand for their human capital and awareness
levels, respectively. Lastly, a few patients' health care characteristics are considered which directly pertain to his/her
health seeking behavior and acuteness of the morbidity. Time needed to reach the hospital, length of stay (LOS) at
the hospital (proxy for severity of the illness), and existence of earlier morbidity within a year prior to this episode
are considered. Certain variables are transformed with functions to correct the skewness in data following ladder
analysis in STATA 12.

3 | RESULTS

Table 2 posits the breadth of the policy to bring the patients within the net of the FPMS program is around 64.51%
on the average, with non‐poor visiting FPMS in far higher shares (difference being statistically significant). Although
this rate of coverage is not actually unusually low, there is still large possibility of enhancing the coverage. Coming to
the depth of services, the picture is varied: while around 16% of the patients did not buy any medicines, (although
overwhelmingly most of the medicines were listed in the essential medicine list for FPMS), nearly one‐fourth of
patients visiting FPMS bought all medicines prescribed, representing quite a skewed nature of availability. There is
some difference in buying no medicines across the poor and non‐poor categories. Following the drive to induce
the physicians to prescribe medicines in generic names, nearly 50% of the medicines prescribed were in generics,
with non‐poor patients getting significantly higher shares (Table 2).
Table 3 reports the results of the first‐stage probit regression models, for poor and non‐poor patients separately,
which generate the propensity score of each patient to visit the FPMS. Poor patients appear to be visiting FPMS in
less proportion in a few hospitals outside Kolkata, although there is no such specific tendency in hospitals situated in
Kolkata. Non‐poor patients, however, are far more inclined to visit FPMS in two teaching hospitals situated in the
capital city of Kolkata. Non‐poor cardiology patients visit FPMS significantly less, although, no department or type
of illness control patients to visit FPMS for poorer patients. Share of medicines prescribed in generic forms and
pre‐existence of serious morbidity episodes in households increase the possibility of self‐selecting non‐poor patients
into the program, while belonging to higher caste and access to newspaper both increase that possibility among the
poor patients. Interestingly, the number of medicines needed to be bought (after receiving the free supply from hos-
pitals) strongly determines the possibility of visiting FPMS by the poor people, while that is an insignificant determi-
nant for the non‐poor.
Table 4 presents the summary statistics for patients visiting FPMS and those not visiting FPMS (even when they
needed to buy medicines from outside) both before and after matching by nearest neighbor matching method. The
results posit that most of the means of the indicators of un‐matched controls were considerably different (using t test)
from the corresponding means for patients visiting FPMS, but the difference vanishes after matching. This hints that a
simple comparison of patients visiting and not visiting FPMS is likely to yield biased estimates of their association with
economic outcomes, and we have been successful in reduction of sample bias between the treated and control groups.
The difference in outcome after matching would hence give us the effect of the treatment of the policy itself.
Additionally, estimates of mean bias reported in the last row of Table 4 are less than 15% in both cases of poor
and non‐poor patients, considerably less than the recommended 25% threshold.27 Also, the common‐support restric-
tion does not lead to any significant loss of treated and control individuals (only one out of 446 poor patients and
zero out of 357 non‐poor patients were found out of the common support region). More than 90% controls in the
set of poor households were used less than 4 times with 50% just once, while the corresponding figures for non‐poor
cases were 80% and 56%, thus representing existence of strong common support regions.
DUTTA e563

TABLE 3 Results of probit regression for hospitalized cases

Variable Poor Non‐Poor

Provider level characteristics

Hospital (Balurghat District Hospital ref)
Barasat District Hospital −0.53* 0.04
Barddhaman MC −1.06*** 0.11
Baruipur Sub Divisional Hospital −0.60** −0.37
Bankura Sammelani Medical College Hospital −0.50* 0.01
Chittaranjan Medical College Hospital (Kolkata) 0.28 1.61***
Jalpaiguri District Hospital −0.18 0.17
Krishnanagar District Hospital District Hospital −0.49* 0.19
SSKM Medical College (Kolkata) 0.49 1.15**
Department (Medicine ref)
Pediatrics −0.09 −0.79
Cardiology 0.02 −1.11***
Eye −0.89* 0.24
Orthopedics −0.40 −0.37
Generic share 0.003 0.005**
No of medicines needed to buy 0.05** 0.05
Household level characteristics
Rural (Urban ref) 0.02 −0.25
Caste Reserved (General ref) 0.34*** 0.44
Fuel used Green Fuel (Red ref) 0.06 −0.04
Sanitation Open (Flush/Pit ref) −0.11 −0.26
Patient's education (years of schooling) −0.02 0.03
Access to newspaper Yes (No ref) 0.26* 0.26
Health seeking behavior and morbidity‐related characteristics
Time taken to reach (in min) −0.0005 −0.0007
Earlier morbidity Yes (No ref) 0.02 0.20***
Log of LOS 0.18 0.13
Pseudo R2 0.13 0.17
LR Chi square 82.52*** 81.03***

Source: Analysis of primary data.

Using nearest neighbor matching methods, our results in Table 5 suggest that visit to FPMS is associated with a
significantly lower OOP costs on medicines among non‐poor patients in government hospitals in West Bengal rela-
tive to a set of closely matched control patients, who did opt out from taking any benefits of concessions on MRP.
However, the results are diametrically different for poor patients, for whom the impact of the program initiates a sub-
stantial rise in OOPE for medicine even after matching. Similar opposite results hold for total hospitalization cost
among two groups of patients.

4 | D I S C U S S I O N A N D LE S S O N S L E A R N T

The above analysis suggests that the overall impact of the new policy has been positive primarily for non‐poor
patients visiting public hospitals in West Bengal as the average OOPE has fallen for the treatment group, compared
e564 DUTTA

TABLE 4 Summary statistics for treatment and control (matched and unmatched) of hospitalized patients
Non‐Poor Poor
Mean Mean
Unmatched/
Variable Matched Treated Control t‐Statistic Treated Control t‐Statistic

Provider level characteristics

Hospital code U 0.181 0.02 0.000*** 0.064 0.093 0.000***
M 0.181 0.077 0.498 0.064 0.060 0.848
Department code U 0.307 0.422 0.02** .409 .355 0.10*
M 0.307 0.375 0.237 .409 0.422 0.778
Logarithm of length of stay U 1.51 1.21 0.000*** 1.535 1.273 0.000***
M 1.51 1.49 0.49 1.535 1.538 0.954
Generic share U 59.77 51.58 0.01*** 62.173 60.285 0.465
M 59.77 63.40 0.11 62.173 58.852 0.217
No of medicine to be bought from outside U 5.34 3.92 0.00*** 4.31 3.21 0.00***
M 5.34 5.20 0.09* 4.31 4.43 0.08*
Household level characteristics
Geographical location U 1.46 1.52 0.218 1.72 1.77 0.216
M 1.46 1.48 0.468 1.72 1.72 0.463
Caste U 1.24 1.26 0.67 1.28 1.24 1.10*
M 1.24 1.21 0.16 1.28 1.31 0.61
Fuel used at the household U 1.59 1.38 0.000*** 1.35 1.27 0.04**
M 1.59 1.60 0.10* 1.35 1.35 1.00
Sanitation type U 0.819 0.90 0.02** 1.83 1.90 0.02**
M 0.819 0.824 0.873 1.83 1.86 0.368
Patient's education U 3.63 3.75 0.562 3.66 4.43 0.05**
M 3.63 3.63 0.892 3.66 4.09 0.295
Access to newspaper U 0.517 0.48 0.471 0.54 0.53 0.751
M 0.517 0.52 0.888 0.54 0.55 0.780
Health seeking behavior and morbidity‐related characteristics
Time taken to reach hospital U 90.72 70.56 0.02** 94.71 70.72 0.008***
M 90.72 82.51 0.323 94.71 82.95 0.218
Earlier morbidity U 2.56 1.98 0.000*** 2.56 1.96 0.000***
M 2.56 2.81 0.148 2.65 2.78 0.246
Mean Bias U 18.1 17.4
M 6.5 7.4

Source: Analysis of primary data.

TABLE 5 Outcomes after matching for poor and non‐poor hospitalized patients (Rs)
Non‐Poor Poor
Difference as Difference as
Treated Control % of control Treated Control % of control
Total medicine cost 4089.19 11763.43 −65.24** 3146.99 1230.35 155.78**
Total hospitalization cost 10381.96 17017.25 −38.99** 6488.26 4931.61 31.56***

*Significant at 10% level; **Significant at 5% level. ***Significant at 1% level.

Source: Analysis of primary data.
DUTTA e565

with the control group. This essentially means that the new program has been successful in offering significant finan-
cial protection to the non‐poor patients who received treatments in the government hospitals. It is true in spite of the
fact that coverage and depth of the program appeared to be quite limited. However, the program could not reduce
the OOPE among the poor patients, thus representing a partial failure. The prime reason behind this differential
impact among the two income classes is that only those poor patients tend to visit the FPMS whose number of pre-
scribed medicines is high, and the number received free from hospital is low (thus number of medicines to be bought
from outside is high), while it is actually a weak determinant for the non‐poor to visit the FPMS. Table 6 shows that
within the poor, the treated and control groups had significant difference in receiving free medicines and medicines
to be bought from market outside, although these differences are insignificant among the non‐poor. Thus, the poor
control group has lower number of medicines prescribed and also has low OOPE as they receive a larger number of
prescribed medicines free of cost from the hospital, while the treatment group has higher number of medicines pre-
scribed and has positive (even after huge discount on MRP) OOPE for most of the medicines.
On the other hand, the non‐poor do not care much about the free medicines and visit FPMS only when they
have higher number of medicines prescribed. Also, the poor buy lesser numbers of medicines from the FPMS
(Table 2) compared with the non‐poor, indicating lower chance of reducing the OOPE. Why the former group bought
less medicine from the highly subsidized store is another puzzle. The reason might lie in the backstage as a section of
doctors and nurses constantly warn against bad quality and threaten and force them to return the medicines bought
from FPMS. The anecdotal evidence shows 13.45% and 12.60% patients out of the treatment groups among the
poor and non‐poor categories, respectively, reported that they were forbade by the doctors or nurses to buy medi-
cine and/or they had to return the medicines bought from FPMS and buy again from outside retail shops. The non‐
poor seem to have better awareness and went on to buy medicines from FPMS and/or refused to return them in
spite of pressure. In choosing the providers, a patient needs information about the quality and cost of medical care
offered by all possible and available providers (hospital pharmacy, FPMS, outside retail shops). Further, when the
patients are sent back by doctors to refund the medicine bought from FPMS, there are potential transaction costs
of enforcing treatment agreements (the implicit agency relationship between the patient and the doctor) that might
affect the choice of treatment source or type.
A similar issue that needs special mention here is the fact that the patients for cardiology department appear to
have low possibility of visiting FPMS. Survey on FPMS revealed that the majority of medicines and stents for cardio‐
thoracic operations were available in the FPMS, which meant rational choice of patients for FPMS, given the fact that
the treatment for cardiology is normally expensive. However, the otherwise result probably hints toward induced
demand created by the doctors for buying these medicines and implants from outside stores and the authority needs
to look into the matter more carefully.
Given this discussion, this exercise of evaluation offers several lessons to be learnt for the policy makers. This
newer form of intervention uses the government agency primarily as a regulator and a facilitator, rather than simple
financer or provider of health care. Although this emerges as a great option for the governments, especially those

TABLE 6 Average no of medicines prescribed, received free from hospital, bought from FPMS, and to be bought
outside hospitals among treated and control groups
Poor Non‐Poor

Total no of medicines prescribed Did not go to FPS 5.65 5.77

Went to FPS 6.92 7.36
Received free from hospital Did not go to FPS 2.44 2.37
Went to FPS 2.17 2.08
To be bought either from outside retail shop or from FPS Did not go to FPS 3.22 16.75
Went to FPS 4.72 5.24

Statistical significance (5%) in differences between two consecutive rows of treated and control groups are shown by gray
highlight.
e566 DUTTA

under severe fiscal constraint and with little fiscal space to scale up supply of free medicines, the policy appears to
serve the non‐poor far efficiently. For the poor, the option is always to get the medicines free of charge, rather than
buying them from outside, even at a subsidized cost. Also, the doctors and other care providers seem to act as an
imperfect agent of the government, as they often create a negative image of this scheme as compromise on quality
and induce demand for drugs to be bought from outside (possibly out of vested interests). The poorest, most often
illiterate and unaware, believe them as “life savers” and refrain from visiting FPMS, while non‐poor can somehow
avoid this serious inducement of demand. Thus, for those poor who did not receive most of the medicines free of
cost from hospital, the policy could successfully improve the overall access to medicine, while for the non‐poor it
could protect against financial catastrophe. In this way, the policy had a differentiated impact on two groups of
patients in terms of the cube of policy impact: for the poor, the breadth of coverage improved, and for the non‐poor
the depth of policy could successfully offer financial protection.

5 | C O N CL U S I O N

Creating access to medicines is no doubt a crucial dimension of health policy in developing countries. The key chal-
lenge in reaching this developmental goal is paucity of public funds, coupled with inefficient allocation and usage of
resources. Coexistence of multiple actors, including pharmaceutical industry, wholesalers, retailers, doctors, and med-
ical representatives and their vested interests in different spheres make the problem even more difficult to handle.
The task might get easier if public policy can effectively harness the contributions of major actors in a more synergic
framework. The private sector, by its very nature, seeks opportunities that can maximize its profits. In theory, this
very sector can and should assist the public sector in identifying pathways that can progress toward the essential goal
of access to medicines.
With titanic burden of public debt, the state of West Bengal has attempted to improve the access to medicines
through an extremely innovative and cost‐saving PPP policy initiative to effectively reduce the OOPE of patients.
The uniqueness of the policy has been absence of any public investment in large scale and the fact that benefits
can accrue to patients not only of government hospitals, but also of private clinics. Public sector, usually held respon-
sible for limiting open market competition, intelligently acts to enhance competition among pharmaceutical compa-
nies and retailers to improve the delivery chain.
After 3 years of introduction, the policy is found to offer financial protection primarily to the non‐poor patients
by reducing OOPE. However, this is true only for the relatively better‐off patients, who visited FPMS whenever they
had more medicines per prescriptions and they resisted the pressure of induced demand from the service providers.
For the poorer patients visiting FPMS, the OOPE on medicine, even bought under substantial discounts, were higher
compared with those who did not. The latter group had low OOPE not because they bought the medicines from out-
side retail shops, but because they received them free of cost from hospital pharmacy. Hence, the program failed to
provide financial protection, although they increased the access to medicines significantly. The difference in outcome lay
in the fact that the nature of control groups differed between poor and non‐poor patients. Those poor patients who
did not visit FPMS received most of the medicines free (hence OOPE nearly zero), while for the non‐poor patients
not visiting FPMS bought the drugs from outside retail shops.
Saying this, it must be acknowledged when there are multiple policy interventions available (here free medicine from
hospitals and reduced cost on FPMS), there are serious transaction costs for the patients, especially the poorest and the
most vulnerable. Thus, who self‐select into which policy would depend upon his involvement in other policy and the
co‐founded character between two or more policies make the weakest patients suffer in terms of financial protection.

ACKNOWLEDGEMEN T

This study was funded by the National Rural Health Mission and Department of Health and Family Welfare,
Government of West Bengal, India.
DUTTA e567

The authors have no competing interest.

Our study does not require any human/animal subject to acquire approval from any ethical approval committee.

RE FE R ENC ES
1. The World Medicines Situation Report, 2011. World Health Organization. http://www.who.int/medicines/areas/policy/
world_medicines_situation/wms_intro/en/ . Updated 2011. Accessed June 9, 2015.
2. The World Health Report. Shaping the Future. World Health Organization; 2003 http://www.who.int/whr/2003/en/.
Updated 2003. Accessed June 17, 2015.
3. Velásquez G, Madrid Y, Quick J. Health Reform and Drug Financing: Selected Topic—Health Economics and Drugs, Series No
006. WHO/DAP/98.3. Geneva: World Health Organisation; 1998 http://apps.who.int/medicinedocs/en/d/
Jwhozip33e/. Published 1998. Accessed June 3, 2015.
4. The World Medicines Situation. World Health Organization. http://apps.who.int/medicinedocs/en/d/Js6160e/.
Updated 2004. Accessed June 11, 2015.
5. Whitehead M, Göran D, Timothy E. Equity and health sector reforms: can low‐income countries escape the medical
poverty trap? Lancet. 2001;358(9284):833‐836.
6. Griffin C. User charges for health care in principle and practice (English). Population, Health and Nutrition Department
technical note; no. PHN 8716. The World Bank. http://documents.worldbank.org/curated/en/912001468914691454/
User‐charges‐for‐health‐care‐in‐principle‐and‐practice. Published June 3, 1987. Updated December 2, 2016. Accessed
June 2, 2016.
7. James CD, Hanson K, McPake B, et al. To retain or remove user fees?: reflections on the current debate in low‐ and
middle‐income countries. Appl Health Econ Health Policy. 2006;5(3):137‐153.
8. Wagstaff A, Van Doorslaer E, Van der Burg H, et al. Equity in the finance of health care: some further international
comparisons. J Health Econ. 1999;18(3):263‐290.
9. North DC. Economic performance through time. Am Econ Rev. 1994;84(3):359‐368.
10. Gupta M. Access to medicines and health technologies in India: opportunities and challenges. Paper presented at
BRICS Forum on Health Technology Access among the BRICS Countries: Brazil, Russia, India, China and South
Africa. World Health Organisation India Country Office https://pdfs.semanticscholar.org/presentation/88e5/
6ef0778f705c87ac937e8f4265d7955917d3.pdf. Updated October 17, 2017. Accessed October 9, 2017
11. Planning Commission of India. High Level Expert Group Report on Universal Health Coverage for India. New Delhi:
Government of India; 2011.
12. Ghosh S. Catastrophic payments and impoverishment due to out‐of‐pocket health spending. Econ Pol Wkly.
2011;46(47):63‐70.
13. Garg C, Karan AK. Reducing out‐of‐pocket expenditures to reduce poverty: a disaggregated analysis at rural‐urban and
state level in India. Health Policy Plan. 2009;24(2):116‐128.
14. Dutta A, Bandyopadhyay S, Ghosh A. Access to Medicines & Public Hospitals and Management Issues. In: Gurtoo A,
Williams C, eds. Developing Country Perspectives on Public Service Delivery. India: Springer; 2014.
15. World Health Organisation. Supply of Free Essential Medicines to Achieve Universal Health Coverage in India. New Delhi:
India Country Office; 2013.
16. Selvaraj S, Mukhopadhyay I, Kumar P, et al. Universal access to medicines: evidence from Rajasthan, India. WHO
South‐East Asia Journal of Public Health. 2014;3(3–4):289‐299.
17. Poverty Estimates for 2004‐05. Press Information Bureau. Planning Commission of India. Government of India, 2007.
http://planningcommission.gov.in/news/prmar07.pdf. Released March 2007. Accessed July 3, 2015.
18. Dutta A, Bose M. Heath equity in access and benefit incidence: a tale of two states in India. International Growth Centre
LSE http://www.theigc.org/wp‐content/uploads/2015/01/Dutta‐Bose‐2014.pdf. Published 2014. Updated 2015.
Accessed August3, 2015.
19. Establishment of fair price outlet for medicines, consumables and implants at selected government hospitals through
public private partnerships (PPP), 2012. Government of West Bengal. http://www.wbhealth.gov.in. Updated 2012
Accessed December 12, 2015.
20. Leach B, Paluzzi J, Munderi P. Prescription for healthy development: increasing access to medicines. UN Millennium
Project on Task Force on HIV/AIDS, Malaria, TB and Access to Essential Medicines. London: Earthscan; 2005.
21. Dutta A, Bandyopadhyay S. Access to Medicine in West Bengal: An Evaluation. Kolkata: Department of Health and Family
Welfare. India: Government of West Bengal; 2016.
e568 DUTTA

22. Planning Commission of India. Report of the Expert Group to Review the Methodology for Measurement of Poverty.
New Delhi: Government of India; 2014.
23. Khandekar S, Koolwal G, Samad H. Handbook on Impact Evaluation: Quantitative Methods and Practices. Washington, DC:
The World Bank; 2010.
24. Dehejia R, Wahba S. Propensity score matching methods for non‐experimental causal studies. Rev Econ Stat.
2002;84(1):151‐161.
25. Htet S, Alam K, Mahal A. Economic burden of chronic conditions among households in Myanmar: the case of angina and
asthma. Health Policy Plan. 2014;30(9):1173‐1183.
26. Health Financing for Universal Coverage. World Health Organization. http://www.who.int/health_financing/strategy/
dimensions/en/. Updated 2011. Accessed June 5, 2017.
27. Ho D, Imai K, King G, Stuart E. Matching as non‐parametric preprocessing for reducing model dependence in parametric
causal inference. Political Analysis. 2007;15(03):199‐236.

How to cite this article: Dutta A, Bandyopadhyay S. Policy intervention for access to medicine: Does it work
similarly for poor and non‐poor? Int J Health Plann Mgmt. 2019;34:e557–e568. https://doi.org/10.1002/
hpm.2671