Resource shortage in public health facilities and private pharmacy practices in Odisha, India

Abstract

In low- and-middle-income countries (LMICs), private pharmacies play a crucial role in the supply of medicines and the provision of healthcare. However, they also engage in poor practices including the improper sale of medicines and caregiving beyond their legal scope. Addressing the deficiencies of private pharmacies can increase their potential contribution towards enhancing universal health coverage. Therefore, it is important to identify the determinants of their performance. The existing literature has mostly focused on pharmacy-level factors and their regulatory environment, ignoring the market in which they operate, particularly their relationship to existing public sector provision. In this study, we fill the gap in the literature by examining the relationship between the practices of private pharmacies and resource shortages in nearby public health facilities in Odisha, India. This is possible due to three novel primary datasets with detailed information on private pharmacies and different levels of public healthcare facilities, including their geospatial coordinates. We find that when public healthcare facilities experience shortages of healthcare workers and essential medicines, private pharmacies step in to fill the gaps created by adjusting the type and amount of care provision and medicine dispensing services. Moreover, the relationship depends on their location, with public facilities and private pharmacies in rural areas performing substitutive caregiving roles, while they are complementary in urban areas. This study demonstrates how policies aimed at addressing resource shortages in public health facilities can generate dynamic responses from private pharmacies, highlighting the need for thorough scrutiny of the interaction between public healthcare facilities and private pharmacies in LMICs.

Key messages.

• This study examines the association between the practices of private pharmacies and the resource shortage in public health facilities in Odisha, India.

• We find that the availability of health workers and essential medicines in public health facilities are important determinants of medicine dispensing and care provision by private pharmacies, but the relationships depends on their location.

• Public facilities and private pharmacies in rural areas perform substitutive caregiving roles, while in urban areas they perform complementary caregiving roles.

• The findings of this study highlight how addressing resource shortage at public health facilities can improve private pharmacies in India, a key provider of medicines and health care in the country.

Introduction

Private pharmacies play an important role in the supply of medicines and the provision of health care services in low- and-middle-income countries (LMICs), often serving as a first point of contact with the health system (Smith, 2009; Sudhinaraset et al., 2013; Miller and Goodman, 2016; Mor et al., 2023). A recent survey in Odisha, India, revealed that 17% of respondents who reported being ill sought care at private pharmacies, while 65% purchased drugs from pharmacies (Haakenstad et al., 2022; Kalita et al., 2023). The dominance of private pharmacies in health systems stems from their advantages, such as easy accessibility, convenient hours, medicine availability, personal familiarity, and perceived confidentiality (Miller and Goodman, 2016; Lamba et al., 2021; Kalita et al., 2023). However, they also engage in poor practices including the improper sale of medicines and caregiving, often beyond their legal scope. These practices are undertaken without appropriate treatment guidelines adherence, without referring clients to a qualified doctor when necessary, or without adequate consultation or counselling with clients (Smith, 2009; Miller and Goodman, 2016; Lamba et al., 2021).

Addressing these deficiencies can enhance private pharmacies’ potential to contribute to advancing universal health coverage (UHC) in LMICs. A crucial step towards improving private pharmacies is to identify the factors that influence their practices and performance. While a few studies have examined the behaviour of private pharmacies, they focus predominantly on pharmacy characteristics and the broader regulatory environment in which they operate (see Miller and Goodman, 2016). However, the analyses in most studies fail to consider private pharmacies as part of a larger system that interact with a variety of formal and informal health providers, including public and private healthcare facilities, laboratories, as well as other pharmacies. Neglecting these interactions hinders a more complete understanding of the factors that influence pharmacy practices.

The objective of this study is to investigate the influence of resource deficiencies in public healthcare facilities on the practices of private pharmacies. Resource deficiencies in public healthcare facilities are widespread in LMICs (Hinman et al., 2006; World Health Organization, 2006; Liu et al., 2017; Phuong et al., 2019). Knowing the relationship between public sector resource deficiencies and private pharmacy practices enables us to appraise whether strategies directed at strengthening public sector health facilities affect private pharmacies in LMICs with mixed markets for pharmacies.

Specifically, our study aims to empirically examine the associations between the practices of private pharmacies and the extent of resource shortages in public healthcare facilities in Odisha, India. We assess whether medicine dispensing and care provision practices of private pharmacies are related to the availability of health workers and essential medicines in public healthcare facilities. Odisha, a state in eastern India with a population of over 41 million (Registrar General of India, 2011), is among the six most impoverished states in the country, making the findings of this paper relevant for similar states with low development indices.¹ We utilize three novel datasets containing detailed information on public primary care facilities, public secondary and tertiary care facilities, and private pharmacies. The availability of geospatial coordinates of pharmacies and facilities in the data is especially crucial since it enables us to link public healthcare facilities and private pharmacies located close by.

The pharmacy practices considered in this study are the sale of medicines, the primary function of pharmacies and the provision of medical care. For the sale of medicines, we consider the sale of branded medicines and branded generic medicines. In India, branded medicines are manufactured by prominent Indian or multinational companies while the cheaper branded generics are generic equivalents sold under a brand name to boost customer recognition (Kotwani, 2013). While private pharmacies mostly sell branded or branded generic medicines, public healthcare facilities procure unbranded generic medicines with non-proprietary name labelling (Aivalli et al., 2018). Currently, there are no laws in the country to ensure the prescription and dispensing of generic medicines. While there are recommendations for physicians and pharmacists to promote generic medicine usage, these are largely ignored (Roy and Rana, 2018; Perappadan, 2019),² resulting in catastrophic health expenditures for Indian households (Haakenstad et al., 2022).

We analyse three activities of medical care provision: the advice provided by pharmacies regarding medication usage, the administering of injections and intravenous fluids (IVs). According to the Pharmacy Act of 1948 and the Pharmacy Practice Regulations of 2015, only registered pharmacists at private pharmacies can dispense certain classes of drugs with a valid prescription from a registered allopathic doctor. Pharmacies are not legally permitted to administer injections and IV fluids.³

Our measure of resource shortages is the availability of health workers and essential medicines in public facilities. These shortages have been found to be substantial. For example, in 2019–20, there were shortages in general physicians and specialists of between 7 and 76% relative to national requirements (Government of India (GoI), 2020; Karan et al., 2019; Rao et al., 2011; Yip et al., 2022). Shortage of essential medicine in public facilities is also widespread (Kotwani et al., 2007; Cameron et al., 2009; Prinja et al., 2015; Chebolu-Subramanian and Sundarraj, 2021; Yip et al., 2022). Furthermore, healthcare resources are unequally distributed across the country, particularly between rural and urban areas (Rao et al., 2011; Selvaraj et al., 2014; Karan et al., 2019).

This paper contributes to three strands of related literature. First, it provides quantitative evidence on the practices of 511 private pharmacies across six districts in one state in India. While pharmacy practices in the country have been documented previously, most existing studies suffer from very small sample sizes (ranging from 12 to 75 pharmacies) and are limited in their geographical coverage (Miller and Goodman, 2016; Barker et al., 2017). Our study contributes specifically towards understanding the extent of illegal pharmacy practices of administering injections and IV fluids which have not been previously studied.

Second, our study offers the first insight into the association between the practices of private pharmacies and shortages in public sector facilities in India. Only one Ugandan study has explored this theme (Fitzpatrick, 2022). Another related study to ours describes the concentration of private pharmacies around healthcare facilities in Ujjain district in India, without explicitly analysing a relationship between the two providers (Sabde et al., 2011). Such associations are likely, since resource shortages in public healthcare facilities are expected to lead to a lower number of patient visits to these establishments. Patients would, instead, likely self-medicate or seek primary care services from the private sector, including private pharmacies, or purchase drugs from private pharmacies after a public sector consultation (Haakenstad et al., 2024). This is especially so as the quality of care dispensed by trained and untrained healthcare providers in India is minimal (Das and Hammer, 2007). Profit maximizing motive of private pharmacies, unmet healthcare needs of patients and the lax regulatory environment in India are likely to incentivize pharmacies to increase the number of patients they offer care and the volume of non-generic medicines sold.⁴

Third, this paper contributes to the discussion around the interactions between public and private healthcare sectors (Das, 2011; Fitzpatrick, 2022; Bedoya et al., 2023; Lim, 2023; Haakenstad et al., 2024) as well as in other product and service markets (Jiménez Hernández and Seira, 2022; Andrabi et al., 2023).⁵ A systematic review of public and private healthcare systems in LMICs found that private providers had better service quality (e.g. shorter waiting times) but poorer technical quality (e.g. provider competency) outcomes (Basu et al., 2012; Morgan et al., 2016). Studies on consumer behaviour have examined the role of quality gap between public and private providers plays in shaping the choice of providers (Costa-Font and García, 2003; Barik and Desai, 2014; Rout et al., 2019). On the supply side, there is a lack of empirical evidence on how private providers decide on which products and services to offer, and how their decisions are influenced by public sector quality. While the performance of the private sector is expected to be intrinsically linked to the performance of the public sector, the relationship is not well understood. Yet this knowledge is essential in appraising the role the private sector plays in the health systems of LMICs, particularly in its role in achieving UHC (Morgan et al., 2016).

Materials and methods

Study context

Private pharmacies are an important player in the Indian health system. According to an Indian association of pharmacists, there are more than 850 000 registered private retail pharmacies in the country (All India Organisation of Chemists and Druggists (AIOCD), 2021). This is an underestimate since many shops sell medicines but are not registered as pharmacies (Kalita et al., 2023).

Private pharmacies in the country function alongside a highly heterogeneous private health sector, comprising of super-speciality internationally accredited hospitals to traditional healers and a three-tier public health system. The first point of contact of patients with the public healthcare system is the sub-centre (SC) which is required to be staffed by healthcare workers to provide preventive, promotive and basic curative care services. Next are the primary health centres (PHCs), the lowest public healthcare facility required to be staffed by a medical doctor.⁶ Community Health Centres (CHCs) constitute the secondary level of healthcare and are intended to serve as a referral centre for four PHCs and to provide obstetric and specialist healthcare. Patients from these lower-level facilities are designed to be referred to sub-district, district, public medical college hospitals, or other government hospitals (Government of India (GoI), 2012).

Data and measures

The data for this study are based on surveys of three types of healthcare providers in Odisha: public primary healthcare facilities, public secondary and tertiary healthcare facilities and private pharmacies. Data collection was carried out in six districts across the state between August 2019 and March 2020. Districts were chosen based on administrative regions and level of development. From these six districts, 30 blocks (the sub-district administrative unit in India) were selected randomly using probability-proportional-to-size sampling such that the more populous districts had more blocks in the sample. The location of the 30 sampled blocks in Odisha, demarcated in dark blue, is shown in Figure 1.⁷

Figure 1.

Map of Odisha showing the sample blocks

Notes: The dark blue indicates the 30 blocks that were sampled in the 6 districts where data collection occurred. The red dots indicate the interviewed private pharmacies. The light blue indicates blocks in the 6 districts that were not included in our analysis since data on public primary facilities was not collected from these.

Different data collection designs were used for each of the three provider surveys. A census of all public secondary and tertiary healthcare facilities in the 30 sampled blocks across the six districts was identified using government records. For the public primary healthcare facilities, a census of all PHCs in the 30 blocks was undertaken. From these sampled blocks, SCs located in or near villages or wards selected for inclusion in a corresponding household survey were also identified.⁸ Through the household survey, we further identified SCs by asking respondents to indicate their preferred providers for outpatient care.

Our final sample comprised 122 public secondary and tertiary healthcare facilities and 396 public primary healthcare facilities. This corresponds to an overall response rate of 78%, with non-response primarily due to participation refusal or interruptions due to COVID-19. Data on services offered, staffing, infrastructure, and other topics were collected through questionnaires administered to the officers-in-charge of the facilities, and, where required, to other staff members. Supplementary Appendix Table A1 lists the distribution of the surveyed public healthcare facilities across the six districts.

In addition to public facilities, we collected detailed information from 1021 private retail pharmacies by surveying store owners, managers, or other employees available at the time of the interview.⁹ Since there is no reliable repository of private pharmacies in Odisha, we utilized information from multiple sources to draw the sample. First, an enumeration list of private pharmacies was created using information on households’ preferred providers for outpatient care obtained from the household survey. The location of these private pharmacies was verified and mapped by the enumerators. Second, all private pharmacies located within a 3-km radius of any public healthcare facility were also mapped. Note that the pharmacies identified from these two sources could be located in any block within the six sampled districts. Lastly, private pharmacies were identified through snowball sampling, where already interviewed healthcare providers were asked to name private pharmacies in the study locations for inclusion in the survey.¹⁰

We excluded one block because of the lack of within-block variation in the independent variables to allow for block-level fixed-effects elaborated further in the article. After excluding pharmacies with missing information on pharmacy characteristics, we arrived at a sample of 612 private pharmacies.

Using the information collected from private pharmacies, we have five indicators of their dispensing and care provision practices in a typical week. These are the sale of branded medicines as a share of the total medicines sold, the sale of branded generic medicines as a share of the total medicines sold, the number of customers that pharmacy staff advised on what medicines to take, the number of customers that the pharmacy administered injections, and the number of customers that the pharmacy administered IV fluids.

The two key independent variables are the availability of healthcare workers and essential medicines in public facilities. The availability of healthcare workers is measured as the percentage of sanctioned healthcare worker positions that are filled at a facility, with the numbers of sanctioned and filled positions being self-reported by the respondents. This indicator is used in government estimates of medical personnel shortages at the facility level, as opposed to region-level shortages measured relative to the population density (Government of India (GoI), 2022). The healthcare workers considered in this study are doctors, nurses, mid-level providers, paramedics and other clinical workers.

Following the literature on medicine shortages in India, the availability of essential medicines is measured as the percentage of select essential medicines in stock at public facilities (Kotwani et al., 2007; Cameron et al., 2009; Prinja et al., 2015). This variable is based on interviewer records after examining whether facilities had a stock of 41 frequently used primary care essential medicines on the Odisha Essential Medicine List, chosen through discussions with pharmacists and healthcare providers in the state. The list of essential medicines considered, usually in the form of non-branded generics, is presented in Supplementary Appendix Table A2.

Empirical strategy

To examine the association between private pharmacy practices and resource availability in rural and urban public facilities, we estimate the following regression model for urban and rural pharmacies separately:

(1)

where represents the dispensing and care provision indicators of private pharmacy in block . and are the availability of healthcare workers and essential medicines in public facilities that lie within the boundaries of pharmacy . We will elaborate on these boundaries below. That is, we assume that private pharmacies choose their practices based on the location of public facilities as well as their characteristics which are considered exogenous.

is a vector of pharmacy characteristics, including registration status, number of hours the store is open in a week, number of employees, and number of customers the pharmacy gets in a typical week. We also controlled for the number of other private pharmacies within the boundary of pharmacy to capture the interaction of private pharmacies in a market. is a block fixed-effect term that captures unobserved time-invariant block-level characteristics to account for inter-block differences in economic conditions, health status, disease burden, patients’ preferences for quality treatment and facility characteristics. is the error term that captures the impact of all other unobservables that vary across pharmacies, facilities and blocks. The parameters of interest are and

To construct and , it is necessary to first determine which public facilities should count as being close enough to influence the behaviours of private pharmacies. One option is to focus on facilities located in the same administrative unit as the private pharmacy, such as a block.¹¹ However, there are two disadvantages to using official regions. First, administrative boundaries within states do not limit individuals’ choice of healthcare facilities. For example, a patient living close to the boundary of a block may choose to seek care at a hospital in the neighbouring block instead of travelling farther to a hospital within the same block. Second, the area defined within a block is too large to accurately capture the relationship between healthcare facilities.¹²

In view of these considerations, we define geographical boundaries by including public facilities located within a specific straight-line radius from a private pharmacy. This approach is commonly applied in studies of competition in healthcare markets (Garnick et al., 1987; Wright et al., 2016; Longo et al., 2017; Moscelli et al., 2018). The availability of jittered geospatial coordinates of private pharmacies and public facilities in our data allows us to calculate the geodesic distance between pharmacies and facilities.¹³ To determine an appropriate radius, we first assessed the distance between the private pharmacies and the corresponding nearest public facilities. The distribution of the distance is shown in Supplementary Appendix Figure A1, spanning a wide range from 0.03 to 12.5 km, with a median of 1.23 km. In deciding on the optimal radius, a smaller (larger) radius would decrease (increase) the number of pharmacies with one or more public facilities in its boundary. We chose a 3-km radius as the boundary distance for the main regression specification. In all, 90% of the 612 surveyed private pharmacies have one or more public facilities within a 3-km boundary. To assess the sensitivity of our results to the radius choice, we supplement the main analysis by estimating regression models with varying radii. We excluded 63 pharmacies that had no public healthcare facility situated within a 3-km boundary, reducing the analysis sample to 549 pharmacies. These facilities were originally included in the sample as preferred pharmacies in the household survey or by already surveyed pharmacies.

Since a private pharmacy can have more than one public healthcare facility within a 3-km radius, we must also determine how to combine a resource availability indicator for the multiple facilities into a single independent variable. Our preferred measure is the mean resource availability across all public facilities within 3 km of a private pharmacy.¹⁴

We estimate Equation (1) using Tobit regression given that the outcomes (e.g. percentage of branded generics, number of customers) we study have a zero-lower bound. Tobit models are used when the outcome variable is a corner solution outcome that takes on the value zero with positive probability but is a continuous variable over strictly positive values (Wooldridge, 2010). As shown in Supplementary Appendix Figure A2, some private pharmacies reported zero for four of the five outcomes, indicating that the pharmacy practices can be seen as a result of two decisions. The first is whether to engage in a service, e.g., administering injections, and conditional on offering the service, the second decision is how many customers to serve. In these applications, using ordinary least squares (OLS) estimation leads to inconsistent estimates, whereas the Tobit model produces coefficient estimates. We also estimate a two-part OLS regression model that explicitly considers the two-stage decision-making using a binary indicator for whether a service is provided and a continuous variable of the number of customers if the service is provided.

We estimate the average marginal effects (AMEs) of resource availability at public facilities on the expected value of the actual outcomes. To test the robustness of our results, we re-estimated Equation (1) using alternative models—fractional regression for the outcomes bound between zero and one (i.e. the shares of branded medicines and branded generic medicines sold) and Poisson regression for count outcomes (i.e. the numbers of customers advised on medicines, administered injections and IV fluids). The Poisson regression, when estimated under pseudo maximum likelihood, is consistent with only the mean being correctly specified even if overdispersion is present (Cameron and Trivedi, 2005) which serves as a sensitivity check of the distributional assumptions underlying the Tobit model. All standard errors are clustered at the district-level to account for the potential correlation of outcomes within a district. We conduct all analyses in STATA 14.2.

Heterogeneity analysis

We examine heterogeneity in the results based on the types of healthcare facilities surrounding private pharmacies. For example, a private pharmacy’s practices may vary depending on whether there is a large sub-district hospital or a small SC located near it. To do so, we consider resource availability at the most dominant public healthcare facility within the 3-km radius of private pharmacies as the independent variable, instead of the mean availability of resources across all nearby public facilities, with dominance being defined by the volume of daily outpatient visits. We then focus on two types of dominant public facilities: public secondary or tertiary healthcare facilities, including hospitals and CHCs, and PHCs. Since the small sample size of PHCs in urban areas and SCs in both areas make the corresponding estimates unreliable, they are not reported. In these three sets of regressions, we rely on district fixed-effects, instead of block fixed-effects, since there are some blocks with only one specific facility type.

Results

Summary statistics

Table 1 shows the characteristics of private pharmacies in our sample. The majority of pharmacies are registered under the Pharmacy Act of 1948 (which regulates the profession and practice of pharmacy in India), though a sizable fraction (29%) is either unregistered or registered under other acts. Pharmacies are open 78 h a week, have an average of 2 employees, with 292 customers in a typical week. There are on average 14 other pharmacies and 2 public facilities located within a 3-km radius of any given pharmacy. Most pharmacies (69%) are in rural areas; these pharmacies have shorter opening hours, smaller staff sizes, fewer customers, and face lesser competition from other pharmacies in the vicinity.

Table 1.

Summary statistics

	(1)	(2)	(3)
	All	Rural	Urban
	Outcome variables
Number of customers advised on medicine use	87.91	86.96	90.11
	(123.9)	(120.5)	(131.7)
Number of customers given injections	9.069	11.34	3.819
	(27.91)	(31.18)	(17.23)
Number of customers administered IVs	4.260	5.836	0.627
	(21.95)	(26.07)	(2.641)
Share of branded medicines sold	0.710	0.717	0.695
	(0.301)	(0.305)	(0.293)
Share of branded generics sold	0.0938	0.0937	0.0939
	(0.138)	(0.138)	(0.137)
	Main independent variables
% of sanctioned positions on duty across all facilities	70.70	76.03	58.40
	(26.38)	(26.01)	(22.94)
% of essential medicines in stock across all facilities	75.89	81.12	63.80
	(19.80)	(17.70)	(19.15)
% of sanctioned positions on duty in the dominant facility	59.24	53.37	72.79
	(27.42)	(25.20)	(27.60)
% of essential medicines in stock in the dominant facility	48.73	44.23	59.10
	(20.95)	(19.48)	(20.59)
	Control variables
Registration: none or other acts	0.293	0.282	0.319
	(0.456)	(0.451)	(0.468)
Registration: Pharmacy Act	0.707	0.718	0.681
	(0.456)	(0.451)	(0.468)
Hours open per week	78.11	76.62	81.54
	(22.21)	(22.81)	(20.43)
Number of employees	2.080	1.849	2.614
	(1.266)	(1.045)	(1.544)
Total number of customers	292.3	256.7	374.4
	(266.0)	(204.3)	(358.7)
Number of other pharmacies in a 3-km radius	13.51	6.337	30.07
	(16.87)	(6.840)	(21.02)
Number of public facilities in a 3-km radius	2.224	2.376	1.873
	(1.529)	(1.683)	(1.016)
Rural	0.689
	(0.460)
Observations	549	383	166

Note: SD in parentheses.

Moving to pharmacy practices, roughly 88% of customers were offered advice on medicine use while the number of customers being administered injections or IV fluids is substantially lower. The proportion of customers receiving injections and IVs in rural areas is notably higher compared with urban areas. Branded medicines accounted for 71%, the largest share of a pharmacy’s total medicine sales; branded generics accounted for 9%, reflecting the smaller scale of the generic medicine market in Odisha. On resource availability, public facilities have on average less than three-quarters of their sanctioned staff positions filled, with staff shortage being a larger issue in urban facilities (58 vs 76%). The reverse is observed for essential medicine stock, with rural facilities experiencing more significant shortfalls compared with urban facilities (64 vs 81%).

Regression results

The estimated associations between resource availability and pharmacy practices are summarized in Table 2 where we show the estimated AME from the Tobit regressions using the mean resource availability and a boundary of 3-km. The estimated coefficients for all independent variables are reported in Supplementary Appendix Table A3. For rural pharmacies (Panel A), we find some evidence indicating that care provision practices by private pharmacies are associated with resource availability in public facilities. Specifically, the number of customers given injections is negatively associated with the availability of healthcare workers and essential medicines in rural public facilities. A one percentage point (pp) increase in the share of sanctioned staff positions decreases the number of customers administered injections at a rural pharmacy by 0.2, a 1.8% decrease (0.2/11.34) in the number of customers administered injections at pharmacies, given the mean value of 11.34 (Table 1). Similarly, a one pp increase in the share of essential medicines in stock at the nearby public facility decreases the number of customers administered injections at the pharmacy by 0.21, i.e., a 1.9% (0.21/11.34) decrease. Increased availability of essential medicines in nearby public facilities in rural areas also decreased the number of customers administered IV fluids by 2.4% (=0.14/5.8).

Table 2.

AME of resource availability across all public health facilities on pharmacy practices

	(1)	(2)	(3)	(4)	(5)
	Customers advised on medicine use	Customers given injections	Customers administered IV fluids	Branded medicine share	Branded generics share
	A: Rural
Availability of workers	0.084	−0.198***	−0.052	−0.000	0.001***
	(0.233)	(0.059)	(0.083)	(0.001)	(0.000)
Availability of medicines	−0.048	−0.209***	−0.137***	−0.000	0.000*
	(0.083)	(0.053)	(0.051)	(0.001)	(0.000)
Observations	383	383	383	383	383
	B: Urban
Availability of workers	−1.015*	0.168***	0.012***	0.003	0.000
	(0.553)	(0.047)	(0.000)	(0.003)	(0.001)
Availability of medicines	−0.135	0.037	0.014***	−0.001	0.001
	(0.178)	(0.023)	(0.000)	(0.002)	(0.001)
Observations	166	166	166	166	166

Note(s): All regressions include block fixed effects and controls for pharmacy characteristics such as registration status, hours the store is open in a week, number of employees, number of total customers in a week and number of other pharmacies in the 3-km circle around the pharmacy. Robust standard errors in parentheses, clustered at the district level.

^***indicates significance at 1% level.

^**at 5% level.

^*at 10% level.

On dispensing practices, the availability of healthcare workers and essential medicines is positively associated with the share of branded generics sold. A one pp increase in the share of sanctioned staff positions filled and the share of essential medicines in stock in rural public facilities, respectively, increase the share of branded generics sold by 1.1% (0.001/0.09).

The estimates for urban pharmacies are shown in Panel B of Table 2 and the results contrast with those of rural pharmacies. The availability of healthcare workers is significantly associated with all three care provision outcomes—advice on medicine use, injections and IVs—though the effects work in different directions. Urban private pharmacies located close to public hospitals with a higher proportion of filled staffing positions advise a lower number of customers on medicine use and administer injections and IVs to a higher number of customers. A one pp increase in the share of sanctioned staff positions in public facilities reduces the number of customers provided with medication advice in a typical week at the pharmacy by 1.02, a reduction of 1.1% (1.02/90.11). Higher staffing availability, on the other hand, also increases the number of customers administered injections and IVs by 4.5% (0.17/3.82) and 1.9% (0.012/0.63). A one pp increase in the share of essential medicines in stock in nearby public facilities also increases the number of customers given IVs at pharmacies by 2.2% (0.014/0.63). On dispensing practices in urban pharmacies, we find that the availability of essential medicines at public facilities does not matter for the sale of medicines for the share of branded generic medicines.

Supplementary Table Appendix A4 presents the results of the two-part regression model. We find that the decision to provide injections and IVs by rural private pharmacies, i.e. the extensive margin decision, is significantly and negatively associated with resource availability at public facilities, similar to the main results. The rural pharmacies’ intensive margin decision of how many customers to serve, conditional on offering the service, is not significantly associated with public facility resource availability. For pharmacies in urban areas, the extensive margin decision to provide injections and IVs along with the intensive margin decision on the number of customers advised on medicine use and administered injections, and the share of branded generics sold are also significantly associated with resource availability at public facilities.

Heterogeneity analysis

Next, we present the results from the Tobit regressions when considering the resource availability in the most dominant public healthcare facility, one with the largest volume of outpatient visits in a day, within the 3-km boundary of private pharmacies, instead of accounting for all facilities in the circle, in Table 3.

Table 3.

AME of resource availability in the dominant public health facilities on pharmacy practices

	(1)	(2)	(3)	(4)	(5)
	Customers advised on medicine use	Customers given injections	Customers administered IV fluids	Branded medicine share	Branded generics share
	Rural: All dominant facilities
Availability of workers	0.399	−0.016	0.026	0.000	0.001**
	(0.317)	(0.029)	(0.021)	(0.001)	(0.000)
Availability of medicines	0.161	−0.120	−0.094	0.000	0.000
	(0.173)	(0.094)	(0.082)	(0.000)	(0.000)
Observations	352	352	352	352	352
	Rural: Higher-level facilities as dominant facilities
Availability of workers	0.582	0.017	0.161**	−0.000	0.000
	(0.399)	(0.074)	(0.079)	(0.001)	(0.000)
Availability of medicines	−0.223	0.124	0.038	0.003	−0.000
	(0.305)	(0.109)	(0.121)	(0.002)	(0.001)
Observations	164	164	164	164	164
	Rural: PHCs as dominant facilities
Availability of workers	0.548	0.092	0.102	−0.000	0.000
	(0.459)	(0.170)	(0.087)	(0.006)	(0.002)
Availability of medicines	0.034	0.071	0.046	0.004*	0.000
	(0.159)	(0.058)	(0.036)	(0.002)	(0.001)
Observations	132	132	132	132	132
	Urban: All dominant facilities
Availability of workers	−0.429***	0.233***	0.023***	0.004	−0.000
	(0.087)	(0.087)	(0.001)	(0.003)	(0.001)
Availability of medicines	0.323**	0.022	0.010***	0.002	0.001***
	(0.139)	(0.024)	(0.000)	(0.001)	(0.000)
Observations	159	159	159	159	159
	Urban: Higher-level facilities as dominant facilities
Availability of workers	−1.367**	0.034	−0.010***	0.002	0.000
	(0.611)	(0.053)	(0.002)	(0.003)	(0.001)
Availability of medicines	0.912	0.049	0.034***	0.003	0.003***
	(0.769)	(0.045)	(0.007)	(0.003)	(0.001)
Observations	136	136	136	136	136

Note(s): Hospitals include medical colleges, district hospitals, sub-district hospitals and other government hospitals. We do not consider SCs separately because of their small sample sizes. There is only 1 PHC in an urban area in our sample. All regressions include block fixed effects and controls for pharmacy characteristics such as registration status, hours the store is open in a week, number of employees, number of total customers in a week and number of other pharmacies in the 3-km circle around the pharmacy. Robust standard errors in parentheses, clustered at the district level.

^***indicates significance at 1% level.

^**at 5% level.

^*at 10% level.

We find that in urban areas the relationships between resource availability in dominant facilities, especially for hospitals or CHCs, and the practices of private pharmacies are similar to the main results where mean resource availability across all public facilities near private pharmacies was used as the independent variables. In rural areas, the results are heterogeneous depending on the type of public facility that is the largest in the 3-km boundary around private pharmacies.

As shown in Panel A of Table 3, in rural areas, the healthcare workers in the dominant public facilities is the only resource that continues to have a significant and positive relationship with the share of branded generics sold. When we look at the different types of dominant facilities (Panels B and C), we note that the main results reported in Table 2 no longer hold, irrespective of the facility type. Instead, when rural pharmacies have high-level facilities as the dominant facility, the availability of healthcare workers is significantly and positively associated with the number of customers given IV fluids. When PHCs are the dominant facility, the availability of essential medicines significantly increases the share of branded medicines sold at nearby private pharmacies.

In urban areas (Panel D of Table 3), a new finding is that the availability of essential medicines in stock in the nearby dominant facility increases the number of customers advised on what medicines. However, the effect is very small—a one pp increase in the share of essential medicines in stock in dominant public facilities is associated with a 0.4% in the number of customers advised in a typical week. For urban pharmacies with higher-level public facilities as the dominant ones, most coefficients are similar to the main results in terms of direction and significance.

Sensitivity checks

As mentioned earlier, we test the robustness of the main results by re-estimating Equation (1) using Poisson regression for the customer numbers and fractional regression (probit) for the share of medicine sales. The AMEs from the Tobit regression are repeated in Panels A and C of Table 4 for comparison while the AMEs using the alternate estimations are presented in Panels B and D. The alternate AME estimates have the same signs as the Tobit estimates and their statistical significance are similar but with a few exceptions. For example, the estimated coefficients on IV injections across the two models have the same sign and magnitude (for worker availability), with larger standard errors from the Poisson model. We do, however, observe differences in the magnitudes of the estimated AMEs in the two panels.

Table 4.

AME of resource availability across all public health facilities on pharmacy practices using alternate regression models

	(1)	(2)	(3)	(4)	(5)
	Customers advised on medicine use	Customers given injections	Customers administered IV fluids	Branded medicine share	Branded generics share
	Rural
	Panel A: Tobit
Availability of workers	0.084	−0.198***	−0.052	−0.000	0.001***
	(0.233)	(0.059)	(0.083)	(0.001)	(0.000)
Availability of medicines	−0.048	−0.209***	−0.137***	−0.000	0.000*
	(0.083)	(0.053)	(0.051)	(0.001)	(0.000)
	383	383	383	383	383
	Panel B: Poisson or Fractional regression
Availability of workers	−0.013	−0.193	0.016	0.000	0.001***
	(0.385)	(0.149)	(0.151)	(0.001)	(0.000)
Availability of medicines	0.066	−0.099	−0.043	−0.000	0.001
	(0.080)	(0.094)	(0.045)	(0.001)	(0.000)
Observations	383	383	383	383	383
	Urban
	Panel C: Tobit
Availability of workers	−1.015*	0.168***	0.012***	0.003	0.000
	(0.553)	(0.047)	(0.000)	(0.003)	(0.001)
Availability of medicines	−0.135	0.037	0.014***	−0.001	0.001
	(0.178)	(0.023)	(0.000)	(0.002)	(0.001)
	166	166	166	166	166
	Panel D: Poisson or Fractional regression or Poisson
Availability of workers	−1.138**	0.201	0.013	0.003	0.000
	(0.566)	(0.137)	(0.012)	(0.002)	(0.001)
Availability of medicines	0.196	−0.116**	0.009***	−0.001	0.001
	(0.223)	(0.045)	(0.001)	(0.002)	(0.001)
Observations	166	166	166	166	166

Note(s): All regressions include block fixed effects and controls for pharmacy characteristics such as registration status, hours the store is open in a week, number of employees, number of customers in a week and number of other pharmacies in the 3-km circle around the pharmacy. Robust standard errors in parentheses, clustered at the district level.

^***indicates significance at 1% level.

^**at 5% level.

^*at 10% level.

We replicate the analyses by using two different radii—2 and 4 km—to identify public facilities whose resource availability could influence the practices of private pharmacies. As shown in panels B and C of Table 5, whether resource availability at public facilities located within the two circles matters for private pharmacies in rural areas depends on the outcomes considered. Panels C, E and F indicate that the results are largely consistent with the main findings when the radius is larger in rural and urban areas.

Table 5.

AME of resource availability across all public health facilities on pharmacy practices using different radii

	(1)	(2)	(3)	(4)	(5)
	Customers advised on medicine use	Customers given injections	Customers administered IV fluids	Branded medicine share	Branded generics share
	Rural: 3 km
Availability of workers	0.084	−0.198***	−0.052	−0.000	0.001***
	(0.233)	(0.059)	(0.083)	(0.001)	(0.000)
Availability of medicines	−0.048	−0.209***	−0.137***	−0.000	0.000*
	(0.083)	(0.053)	(0.051)	(0.001)	(0.000)
Observations	383	383	383	383	383
	Rural: 2 km
Availability of workers	0.379*	−0.144	−0.046	0.001	0.000
	(0.197)	(0.090)	(0.076)	(0.001)	(0.000)
Availability of medicines	−0.062	−0.111*	0.011	−0.000	0.000
	(0.107)	(0.059)	(0.069)	(0.001)	(0.000)
Observations	335	335	335	335	335
	Rural: 4 km
Availability of workers	0.051	−0.159*	−0.063	−0.000	0.001*
	(0.161)	(0.094)	(0.078)	(0.001)	(0.000)
Availability of medicines	−0.149	−0.096	−0.185***	0.001	0.001
	(0.143)	(0.071)	(0.046)	(0.001)	(0.000)
Observations	362	362	362	362	362
	Urban: 3 km
Availability of workers	−1.015*	0.168***	0.012***	0.003	0.000
	(0.553)	(0.047)	(0.000)	(0.003)	(0.001)
Availability of medicines	−0.135	0.037	0.014***	−0.001	0.001
	(0.178)	(0.023)	(0.000)	(0.002)	(0.001)
Observations	166	166	166	166	166
	Urban: 2 km
Availability of workers	−0.066	0.101*	0.009***	0.001	0.000
	(0.089)	(0.060)	(0.001)	(0.001)	(0.001)
Availability of medicines	−0.312*	0.019	0.007***	−0.001	0.001
	(0.161)	(0.036)	(0.001)	(0.002)	(0.001)
Observations	148	148	148	148	148
	Urban: 4 km
Availability of workers	−0.226	0.227**	0.028***	0.002	−0.000
	(0.177)	(0.099)	(0.001)	(0.002)	(0.001)
Availability of medicines	0.120	0.198***	0.033***	−0.004***	0.002***
	(0.375)	(0.053)	(0.001)	(0.001)	(0.000)
Observations	181	181	181	181	181

Note(s): All regressions include block fixed effects and controls for pharmacy characteristics such as registration status, hours the store is open in a week, number of employees, number of customers in a week and number of other pharmacies in the 3-km circle around the pharmacy. Robust standard errors in parentheses, clustered at the district level.

^***indicates significance at 1% level.

^**at 5% level.

^*at 10% level.

The number of pharmacies in the 4-km circle is less than the number in the 3-km circle because we dropped an additional block due to the lack of within-block variation in the independent variables to allow for the inclusion of block-level fixed effects.

Discussion

In this study, we examined the relationship between the dispensing and care provision practices of private pharmacies and the availability of healthcare workers and essential medicines in public health facilities. Using novel data from 511 private pharmacies and public facilities located within a 3-km boundary of the pharmacies in Odisha, we find that resource availability at public facilities is an important determinant of four of the five pharmacy practices considered in this study: the volume of patients advised on what medicines to take, administered injections and IV fluids and the share of branded generic medicines sold by pharmacies in a typical week. Moreover, the relationship between private pharmacies and public facilities depends on their location. However, the sale of branded medicines by private pharmacies is not associated with the supply of healthcare workers or essential medicines at public facilities, irrespective of the location.

These findings suggest differential relationships between public facilities and private pharmacies in rural and urban areas. In rural areas, their relationships are substitutive in nature, since higher-capacity public facilities with fewer resource shortages are associated with fewer customers receiving injections and IV fluids in private pharmacies. In urban areas, public facilities and private pharmacies exhibit more complementary relationships. Higher capacity in public facilities is associated with an increase in the customers administered injections and IV fluids, albeit a decline in private pharmacies providing medical advice. This suggests that patients seek care in public facilities and then go to private pharmacies for IV fluids and injections afterwards.

One potential explanation for these results stems from the fact that the public facilities in rural areas are typically PHCs or SCs, i.e. lower-level facilities that mostly provide outpatient care. These services can be offered by private pharmacies, since employees with pharmacy degrees learn about these as a part of their education (The Hindu, 2009), while others have experience as an apprentice to private doctors (Kamat and Nichter, 1998). On the contrary, in urban areas, the majority of public facilities in urban areas provide secondary and tertiary care, which cannot be replicated by pharmacies. Instead, nearby pharmacies offer complementary services, often following referrals from doctors in higher-level public facilities who have symbiotic financial relations with private pharmacies (Kamat and Nichter, 1998). The results based on the type of dominant facility reflect this explanation.

This study suggests that there exists complex dynamics between public facilities and private pharmacies which can have important implications for access, quality and efficiency of care in India’s healthcare system. On the one hand, private pharmacies can improve access for rural residents where public facilities’ capacities are weak. On the other hand, public facility providers can use private pharmacies to induce unnecessary demand where they stand to gain financially. This can potentially lead to unnecessary financial burden borne by people (Haakenstad et al., 2024). Unfortunately, without assessing the quality of care provided by private pharmacies and understanding the financial ties between public providers and private pharmacies, this study cannot conclude how quality, efficiency and financial risk protection are impacted by the inter-relationships between public facilities and private pharmacies.

Considering the significant role played by private pharmacies, we recommend that, first, policymakers should conduct a thorough and objective assessment of the quality of care provided by private pharmacies, and decide whether regulations, monitoring of standards and training are necessary to ensure that care and advice provided by private pharmacies enhance patient welfare (Shroff et al., 2021; Kalita et al., 2023; Mor et al., 2023).¹⁵ Policymakers may also consider identifying a list of private pharmacies that satisfy certain standards of care and publicize them widely. Second, we recommend that policymakers assess whether IV fluids and injections provided by private pharmacies are clinically necessary and unpack the financial linkages between public providers and private pharmacies. If indeed public providers hold financial interests in private pharmacies and therefore induce unnecessary demand, it will be important to sever these financial links. Otherwise, patients will be bearing huge financial burdens for unnecessary care. Whereas if these IV fluids and injections are necessary and delivered of adequate quality, then policymakers can consider contracting these private pharmacies to substitute where public facilities’ capacities are low. Finally, improving capacities at public facilities, especially in human resources and drug supplies, may help prevent patients from bypassing the public sector to seek care in private pharmacies. These could be achieved by better financial incentives and career development opportunities for the public sector health workforce and by improving supply chains for drugs.

This paper has several limitations. First, due to the absence of a sampling frame for private pharmacies, our sample is not necessarily representative of the universe. However, our reliance on multiple sources to draw our sample for private pharmacies likely minimized the issues. Second, the data from private pharmacies and healthcare facilities are self-reported. For example, we were unable to verify the numbers of customers of the pharmacies and expect our estimates to represent the upper bound of the true statistics. Although the time-invariant sources of reporting bias are removed by the inclusion of block fixed-effects, reporting bias is a limitation of the study. Third, we lack data on private healthcare facilities, which is an important agent in the Indian health system. We are unable to test how the presence of private facilities mediates the relationship between public facilities and pharmacies. Finally, our analysis is descriptive and based on six districts in one state of India; thus caution is required in drawing causal inferences and generalizing the results to all Indian states and other countries.

Despite these limitations, this study provides the first evidence establishing a relationship between private pharmacy practices and resource shortages in public healthcare facilities, and a link between two important pillars of the Indian health system.

Supplementary data

Supplementary data is available at HEAPOL Journal online.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Funding

The study was financially supported by the Bill and Melinda Gates Foundation, and funding for data collection for the study was provided by the Tata Trusts.

Author contributions

W.Y. and B.B. were involved in conception or design of the study.

W.Y., A.K., A.H. and B.B. were involved in data collectionwith implementation by Oxford Policy Management.

B.B. and T.C. were involved in data analysis and interpretation.

B.B. were involved in drafting the article.

B.B., T.C., A.K., A.H. and W.Y. were involved in critical revision of the article.

B.B., T.C., A.K., A.H. and W.Y. were involved in final approval of the version to be submitted.

Reflexivity statement

The authors include four females and one male and span multiple levels of seniority. The corresponding author is a citizen of India with experience in several India-based projects. All authors have extensive research or lived experiences in India or Asia. One of the co-authors has several years of experience working in the Indian health sector; she and the corresponding author have ongoing collaborations in the country. Our authorship team also reflects expertise in several areas including health systems, health & well-being, gender and health economics.

Ethical approval.

Institutional Review Board (IRB) approval for this study was obtained from the Harvard Chan School of Public Health (IRB18-1675), an independent IRB in India, and the health research approval committee of the Odisha Government.

Conflict of interest:

The authors report no conflicts of interest.