RES E AR C H Open Access
Incentives to change: effects of
performance-based financing on health
workers in Zambia
Gordon C. Shen
1*
, Ha Thi Hong Nguyen
2
, Ashis Das
2
, Nkenda Sachingongu
3
, Collins Chansa
2
,
Jumana Qamruddin
2
and Jed Friedman
4
Abstract
Background: Performance-based financing (PBF) has been implemented in a number of countries with the aim of
transforming health systems and improving maternal and child health. This paper examines the effect of PBF on
health workers’ job satisfaction, motivation, and attrition in Zambia. It uses a randomized intervention/control
design to evaluate before–after changes for three groups: intervention (PBF) group, control 1 (C1; enhanced
financing) group, and control 2 (C2; pure control) group.
Methods: Mixed methods are employed. The quantitative portion comprises of a baseline and an endline survey.
The survey and sampling scheme were designed to allow for a rigorous impact evaluation of PBF or C1 on several
key performance indicators. The qualitative portion seeks to explain the pathways underlying the observed
differences through interviews conducted at the beginning and at the three-year mark of the PBF program.
Results: Econometric analysis shows that PBF led to increased job satisfaction and decreased attrition on a subset
of measures, with little effect on motivation. The C1 group also experienced some positive effects on job
satisfaction. The null results of the quantitative assessment of motivation cohere with those of the qualitative
assessment, which revealed that workers remain motivated by their dedication to the profession and to provide
health care to the community rather than by financial incentives. The qualitative evidence also provides two
explanations for higher overall job satisfaction in the C1 than in the PBF group: better working conditions and more
effective supervision from the District Medical Office. The PBF group had higher satisfaction with compensation than
both control groups because they have higher compensation and financial autonomy, which was intended to be part
of the PBF intervention. While PBF could not address all the reasons for attrition, it did lower turnover because those
health centers were staffed with qualified personnel and the personnel had role clarity.
Conclusions: In Zambia, the implementation of PBF schemes brought about a significant increase in job satisfaction
and a decrease in attrition, but had no significant effect on motivation. Enhanced health financing also increased
stated job satisfaction.
Keywords: Performance-based financing, Pay-for-performance, Organizational behavior, Mixed methods, Human
resources for health, Health system strengthening, Zambia
* Correspondence: Gordon.Shen@sph.cuny.edu
1
Department of Health Policy and Management, Graduate School of Public
Health and Health Policy, City University of New York, 55 West 125 Street
Room 806, New York, NY 10027, United States of America
Full list of author information is available at the end of the article
© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to
the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver
(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Shen et al. Human Resources for Health (2017) 15:20
DOI 10.1186/s12960-017-0179-2
Background
Progress towards improving maternal and child health
(MCH) outcomes requires a certain level of human re-
sources to deliver health care services, but this has been
difficult in Zambia due to a human resources for health
(HRH) crisis [1–3]. Zambia faces severe health worker
shortages across all levels of health care, with 93 total
clinical health care workers (HCW)
1
per 100,000 popula-
tion ratio in 2009 [1]. This translates to a 60% gap in the
required versus actual number of clinical health workers
nationwide [4]. An average annual attrition rate of 4%
from 2007 to 2009 effectively cancelled out gains made
in the number of employees recruited [5]. HCWs are
not evenly distributed between rural and urban parts of
the country: 159 clinical health care workers per 100,000
population in urban areas versus 70 clinical health care
workers per 100,000 in rural areas of the country [1].
The HRH shortage has been exacerbated by high levels
of absenteeism (21%), tardi ness (43%), dissatisfaction
(44%), and vacancy (33.5%) rates in 2006 [6]. This situ-
ation is compounded by an imbalance in skill-mix
among HCWs, and limited funding and training institu-
tions. The implementation of HRH Strategic Plans
2006–2010 and Zambian Health Workers Retention
Scheme resulted in increased staff recruitment, appoint-
ment, and retention. However, low salaries and poor
working conditions con tinued to affect health workers’
morale. Workforce maldistribution is further exacer-
bated by brain drain
2
and by increased demands placed
on the health systems by patients with communicable
and non-communicable diseases alike [7, 8].
From 2012 to 2014, the Zambia n government intro-
duced a large-scale performance-based financing (PBF)
program to enhance the performance of existing health
workers for MCH services. PBF programs typically use
incentives to encourage providers to increase the
provision of services and adopt best practices for quality
by following explicit protocols and complying with a sys-
tem of inspection and auditing.
3
Monetary or non-
monetary incentives can be directed at individual health
care workers or at the health facility as a whole, and
therefore districts in Zambia were randomly assigned to
one of three study groups: intervention districts (PBF),
input-based financing districts (C1), and pure control dis-
tricts (C2).
4
To date there is more evidence focused on
the impact of PBF on patient outcomes [9–14] rather than
on health care provider outcomes [15, 16]. This paper
evaluates the effects of this PBF program on health
workers’ satisfaction, motivation, and attrition, and exam-
ines the potential causal pathways leading to such effects.
Theoretical framework
HRH is an important node in the causal pathway from
PBF to desired service provision and ultimately population
health outcomes. Figure 1 is a display of our theory of
change which posits that HRH—at the individual or na-
tional level—changes as a result of the implementation of
a PBF program in Zambia.
5
At the individual worker level,
our model teases apart the type of incentives, as well as
the combination of incentives, that could improve
personnel shortage and low morale. At the national work-
force level, we lay out a set of enabling and disabling con-
ditions that are mediators of PBF and HRH. Introducing
monetary incentives to designated health facilities could,
in theory, he lp achieve systemic objectives to increa se
the availability, distribution, and performance of the
workforce.
In this study we are interested in the differential ef-
fect s of monetary inc entives tied to the activities or ef-
forts of workers (i.e., PBF bonuses) versus alternative
financing modes’ (i.e., enhanced financing, status quo)
effects on two individual HRH outcomes determining
national workforce performance (motivation, job satis-
faction) and an individual HRH outcome (attrition) de-
termining national workforce distribution, as shown in
Fig. 1. Motivation is individuals’ willingness to sustain
efforts towards achieving pre-determined goals. Incen-
tives can be a source of motivation because an individual
or an organization would perform an action in order to
attain a valued resource [ 17]. But health workers may
burnout from increased demands on them to meet PBF
targets in the long run. The empirical evidence of incen-
tives on motivation is mixed,
6
though more recent
empirical evidence from sub-Saharan Africa suggests
that we would expect to see a spike in motivation early
in the PBF progra m’s implementation.
7
We therefore
hypothesize that PBF will have a positive effect on
Zambian health workers’ motivation during our study
period. We further hypothesize that enhanced financing
would also have a positive effect on motivation, but with
a lower magnitude than PBF because enhanced financing
is targeted towards the health facility as a whole rather
than individual health workers and is not linked directly
to performa nce. Similar hypotheses can be made for job
satisfaction.
8
PBF is related to motivation and job satisfaction, which
are predictive of turnover. The first two steps in
Mobley’s heuristic model of employee withdrawal deci-
sion process is evaluation of existing job and experi-
enced job satisfaction or dissatisfaction [18]. Psychology
studies conducted since have found that job satisfaction
was pred icted by the reward and cost values of the job
[19], and that job satisfaction was correlated with job
turnover [20].
9
High levels of mot ivation, like job satis-
faction, reduced the risk of low- and middle-income
country (LMIC) health workers’ intent of leaving their
jobs [21]. The motivation-turnover relationship is medi-
ated by affective commitment [22] and moderated by
Shen et al. Human Resources for Health (2017) 15:20 Page 2 of 15
burnout [23]. These empirical studies that are conducted
either in the lab or in the field cumulative ly suggest that
individual tenure in health care organizations is influ-
enced by extrinsic motivation, and mediated by job satis-
faction with their work setting. We thus hypothesize
that PBF and enhanced financing would each have a
negative effect on turnover, but enhanced financing
would have a lower effect magnitude than PBF because
satisfaction with compensation is a bigger determin ant
of turnover than other aspects of job satisfaction.
The hypothesized magnitude and direction of PBF in-
fluence on HRH are summarized in Table 1.
10
These hy-
potheses reflect the expectations given the organization
behavior literature and features of Zambia’s three-arm
PBF design.
Methods
The study setting (i.e., Zambia) and study intervention
(i.e., PBF in Zambia) are described in Additional file 1.
We gathered quantitative and qualitative data from
health workers and related health centers for this study.
The procedures for linking findings from qualitative and
quantitative research and bringing out their complemen-
tarities can be manifold [24, 25]. Therefore, we chose to
carefully interpret and triangulate the qualitative with
the quantitative data because our aim is model (i.e., Fig. 1)
testing [26].
Study design
This study is part of a broader impact evaluation study
aimed at measuring the effects of PBF on MCH and
other health system outcomes. The evaluation follows a
quasi-experimental design: 30 districts in the country
were triplet-matched on key health systems and out-
come indicators and randomly allocated to each study
arm. Thus there are 10 PBF, 10 C1, and 10 C2 districts.
The district sele ction process, the resulting list of dis-
tricts, their health facilities, and population under study
are further described in Additional file 2.
Health centers in targeted pilot districts were eli-
gible for PBF if it employed at least one qualified
health worker by the end of the first quarter of 2012.
Those health centers received PBF incentive payment s
and emergency obstetric care (EmOC ) equipment.
This P BF agreement is reinforced with an instituti on-
level contract (and a business p lan) signed by DMOs
and health centers, and an individual-level “motivation
contract” signed by health workers and their affiliated
health center. The proportion of the individual PBF staff
bonus to the individual government salary was on average
10% during the entire duration of the project [27].
11
The
determination of health center payment and individual
performance bonuses is further described in Additional
file 3.
Fig. 1 A general conceptual framework on the effects of PBF on HRH
Table 1 Hypothesized magnitude and direction of PBF on HRH
Intervention
(PBF) group
Control 1, or enhanced
financing not conditioned
on outputs
Control 2, or
“business-as-usual”
Motivation ++ +
Job
satisfaction
++ +
Attrition −−
There is a greater magnitude of effect for the intervention group than control
1 group, but the direction should remain the same. Control 2 cells are left
blank because no changes are expected to occur
Shen et al. Human Resources for Health (2017) 15:20 Page 3 of 15
Health centers in the C1 group received additional fi-
nancing intended to equal to the average RBF incentive
payments in intervention districts, as well as EmOC
equipment. This additional financing was not tied to
performance, so health centers spent it as meal allowances
or on rehabilitation of the health center, drugs, outreach
activities, and equipment. Due to administrative bottle-
necks in the financing and procurement processes
adopted by the C1 districts, health facilities in the C1
group received on average a financing amount equal to
56% of the PBF group by the end of the study period.
12
Health centers in C2 group represent “business-as-usual”
since they received neither additional financing nor
EmOC equipment.
This study was supported by the MOH of Zambia.
The research protocol was approved by the Institutional
Ethics Committee of the University of Zambia. Written
informed consent was collected from all respondents.
We kept all personal information confidential, an d no
names were used in the resulting report or journal
articles.
Quantitative data collection and analysis
Quantitative measures pertaining to the HRH outcomes
of interest were assessed through surveys fielded in
health centers at baseline (October–November 2011)
and towards the e nd of the PBF pilot p roject
(September–November 2014).
13
A total of 186 health cen-
ters were surveyed, consisting of 86 in the PBF group, 49
in C1 group, and 51 in C2 group. Up to two health
workers providing MCH services on the day of visit were
interviewed for the survey in every facility, for a total of
683 staff personnel interviewed in two rounds. Statistical
power for the overall evaluation was calculated using
population coverage of services as key outcomes for an
impact evaluation of PBF in Zambia, but power was not
calculated for HRH outcomes in this study.
Motivation and job satisfaction are derived from the
individual worker questionnaire and attrition is based on
the facility assessment. The questions for the motivation
and satisfaction were based on two existing, validated
tools: Minnesota Satisfaction Questio nnaire [28] and Job
Satisfaction Survey [29]. In addition, the variables on
well-being were derived from the WHO Well-Being
Index [30]. The development of the motivation and job
satisfaction constructs are described in more details in
Additional file 4. Attrition was assessed by the number
of authorized staff reported to have left a health center
in the previous 12 months in a health facility survey.
The effects of PBF on key outcome variables were esti-
mated with difference-in-difference framework among
the PBF, C1, and C2 arms for two rounds of data
(baseline and endline). Facility fixed effects analysis was
performed with standard errors clustered at a district
level. District grouping was taken in to account in the
analysis through stratification controls. The difference-
in-difference model can be summarized in the form of a
linear regression equation as follows:
Y
ijtd
¼ γ
0
þ γ
1
PBF
d
þ γ
2
Period
t
þ γ
3
PBF PeriodðÞ
dt
þDP
d
þ X
ijtd
þ ε
ijtd
where Y is the outcome for health worker i under facility
j at time t for district d; γ
0
is a constant; PBF is a binary
variable taking the value of 1 for distric ts in the PBF
treatment area and 0 otherwise; Period is a binary vari-
able where it is 1 for the post-intervention period and 0
otherwise; γ
1
and γ
2
are the coefficients for treatment
and period, respectively; the interaction term is γ
3
which
indicates the difference-in-difference treatment effect;
DP represents the district grouping stratification with a
vector of dummy variables indicating district inclusion
in particular province-level strata; X is a vector of
worker level covariates (age, gender, and staff position); ε
is the random error term. For most of the analysis, pair-
wise comparisons are separately estimated with PBF esti-
mated with the C1 group as the default category, and
then PBF with C2 as the default. The model comparing
C1 with C2 groups is specified exactly the same except
that PBF variable is replaced with a binary variable de-
noting C1. All statistical analyses were done with STATA
version 13.
Results of the three-group comparisons are shown in
Table 2 while results of the two-group compa risons are
shown in Additional file 5. One-way ANOVA shows that
at baseline there was no statistical difference among the
three groups, indicating baseline balance in key charac-
teristics that may mediate the impact of PBF on satisfac-
tion, motivation, and attrition.
Qualitative data collection and analysis
The second objective of our study is to understand the
possible channels through which financial incentives
affect health care providers. The second objective is pur-
sued through in-depth interviews conducted in health
centers, District Medical Offices (DMOs), and provincial
headquarter offices. Interviews were conducted at the
beginning of PBF implementation (“baseline”; November
2011–March 2012) and three years following it (“ end-
line”; January 2015).
14
Organization leaders were inter-
viewed individually, whereas staff members in a similar
level on the organization chart were interviewed in a
group. The sampling goal is to reach theoretical satur-
ation, during which all major concepts are identified and
additional interviews reveal no new information. A total
of 81 interviews were conducted at baseline and 54 in-
terviews were conducted at endline. The interviewees’
demographic information for baseline and endline is
Shen et al. Human Resources for Health (2017) 15:20 Page 4 of 15
shown in Table 3. F4 software was used for transcription,
and NVivo 10 software (QSR International Pty Ltd,
Australia) was used for thematic analysis.
Results
In this section, we present results for the three HRH di-
mensions (motivation, job satisfaction, attr ition), study
group differences for each dimension’s general construct
scores, and for each construct’s constituent variables.
Figure 2 summarizes the intermediary factors that
emerged from interviews, which we will explain along
with the regression analysis results.
Motivation
We did not find support for our hypotheses for any of
the eight motivation constru cts with one exception: re-
spondents in the PBF group reported, out of 100%,
2.42% (p < 0.1) higher on the personal well-being scale
between baseline and endline than those in the C2 group
(Table 4). This aggregate finding is driven by respondents
in the PBF group who felt more calm and relaxed in the
2 weeks prior to reporting between baseline and endline
than those in the C1 group (9.48% higher; p <0.1)orthose
in the C2 group (5.69% higher; p < 0.05). The group differ-
ences for the eight motivation constructs are summarized
in Table 4.
Looking specifically at the individual questions under
each motivation construct (Additional file 6), the PBF
appears to have encouraged staff to willingly give their
time and help each other out when someone fell behind
or had difficulties with his or her w ork; 3.77 % higher
(p < 0.05) between baseline and endline for the PBF
than for the C2 group. Finally, three of the motiv-
ation questions seemed to discern group differences,
which could be us ed and elucidated in future PBF re-
search. The three questions are as follows: “Iwould
prefer to work somewhere else than in this facility”
(12.27% lower between baseline and endline for the
PBF group th an for the C1 group; p <0.1); “My facil-
ity i s a very dynamic and innovative place. People are
willing to take risks to do a job well done” (5.06%
higher between baseline and endline for the PBF
group than for the C2 g roup; p <0.1); and “Following
procedures and rules is very important in my facility”
is 1.95% (p < 0.1) and 4.26% (p < 0.1) higher between
baselineandendlineforthePBFandC1groupwhen
each of them wa s compared with the C2 group.
Interviews, in accordance with the null effects of PBF
on mot ivation, revealed that remuneration alone could
not adequately address two causes of de-motivation:
high workload and low staffing levels. Financial bonuses
paid out by the PBF program were adjusted by workload,
but they were not directly tied to the staffing level. This
Table 2 Mean statistics of workers’ characteristics at baseline and endline in three groups (N = 683)
Variable Baseline Endline
Intervention
(n = 147)
Control 1
(n = 87)
Control 2
(n = 92)
Intervention
(n = 166)
Control 1
(n = 92)
Control 2
(n = 99)
Female 0.42 0.38 0.42 0.41 0.36 0.49
Education-primary 0.06 0.08 0.05 0.04 0.01 0.05
Education-secondary 0.40 0.40 0.30 0.35 0.49 0.27
Education-college 0.52 0.49 0.63 0.60 0.49 0.68
Clinical officer 0.03 0.02 0.04 0.06 0.04 0.03
Nurse 0.25 0.26 0.25 0.33 0.35 0.45
Midwife 0.11 0.13 0.14 0.12 0.09 0.15
Environmental health technicians (EHTs) 0.15 0.09 0.16 0.13 0.08 0.10
Classified daily employees (CDEs) 0.33 0.41 0.32 0.31 0.38 0.22
Other staff 0.67 0.59 0.69 0.69 0.62 0.78
Age 37.43 38.01 36.21 35.82 38.51 35.49
Work-absence 1.20 1.44 1.59 1.12 1.10 1.74
Work-days 5.82 6.26 6.13 6.00 6.24 6.27
Work-hours 51.45 55.90 54.55 52.07 50.33 49.61
Supervision frequency from previous year 4.52 4.32 6.65 5.62 4.58 4.54
Work experience-total 10.06 11.04 9.76 8.03 9.03 7.95
Work experience-current facility 4.55 5.40 4.39 4.27 4.67 5.09
ANOVA test of balance among three groups was performed separately for baseline and endline. Statistical significance is denoted by bold italic (p < 0.01); bold
(p < 0.05); italic (p < 0.1)
Shen et al. Human Resources for Health (2017) 15:20 Page 5 of 15