1 of 28

Evaluation of a national program �to distribute free face masks �for COVID-19 prevention: �evidence from Mbale District in Uganda

Aleksandra Jakubowski, Dennis Egger, Ronald Mulebeke, Pius Akankwasa, Allan Muruta, Noah Kiwanuka, Rhoda Wanyenze

​

2022 Africa Evidence Summit

2 of 28

Outline

  • Background and motivation
  • Methods
    • Setting
    • Design
    • Data
    • Analysis
  • Results
  • Conclusion

1

3 of 28

Background�

2

4 of 28

Background: Masks

  • Key non-pharmaceutical intervention to reduce COVID-19

​

  • Major policy priority in Uganda 2019/2020

​

  • National campaign to distribute masks to all citizens of Uganda >5 years (~30M masks)

​

  • We partnered with Office of Prime Minster, Ministry of Health, and local District officials to evaluate the national distribution campaign in Mbale District in Eastern Uganda

3

5 of 28

Study objectives

​

  • Test whether distribution of free masks affected mask behavior

​

​

  • Test whether pairing mask distribution with education about masks and COVID-19 affected mask attitudes, knowledge, and behavior

​

4

6 of 28

Motivation: Barriers to mask wearing

  • Lack of access to affordable masks?
    • Free distribution improves uptake of health goods (Noor et al. 2007, Ashraf et al. 2010)

​

  • Knowledge of COVID-19 severity or mask effectiveness? (Fetzer et al. 2020)
    • Education about COVID-19 and masks

​

  • Salience of masks?
    • Behavioral nudge to change health behavior (Tversky and Kahneman 1974)

​

  • Social norms?
    • Social signaling can affect adoption of health goods (Karing 2018)

5

7 of 28

Methods�

6

8 of 28

Methods overview

Setting

  • Eastern Region of Uganda
    • Mbale district
  • Study dates: Feb – Apr 2021

​

Intervention

  • Mask distribution
  • Masks + education

​

Data

  • Direct observations of mask behavior
  • Self-reported phone surveys about masks, COVID, socio-economic data

7

9 of 28

Mask distribution campaign & study timeline

8

Distribution �campaign begins

​

Priority to high risk districts: border, highway, dense population

June

2020

Nov

2020

Jan

2021

Feb

2021

Apr

2021

Distribution campaign pauses

​

Holidays & election

Election

Distribution campaign resumes

​

Masks

delivered

to Mbale

Masks delivered to district by national gov’t, last mile logistics up to district

Data collection

10 of 28

Relatively low burden of COVID-19 in Uganda

9

Source: World Health Organization

2020

2021

2022

Uganda

2020

2021

2022

Data collection

Data collection

Elections

Elections

11 of 28

Setting

  • Mbale District
    • Eastern Region of Uganda, near border with Kenya (27 subcounties, 450,000 ppl), includes major trade route
  • Selected 4 subcounties for study
    • Busiu, Busiu Town Council, Bumasikye, Lukhonje (45,000 ppl, 173 villages)
    • Randomly selected 90 villages to collect data
  • Final sample 3 subcounties
    • Busiu, Bumasikye, Lukhonje
    • Busiu TC eliminated pre mask distribution due to implementation issues

​

​

​

10

12 of 28

Intervention

  • Village health teams (VHTs) tasked with distribution

​

  • All villages received masks at the same time

​

  • Volunteers from random villages trained on mask use
    • Information about COVID-19
    • How masks work
    • Proper ways to wear masks
    • Strategies to not forget masks

​

  • VHTs provided script to deliver education to mask recipients

​

11

13 of 28

Data

  • Direct observations
    • Enumerators observed mask use in public places for 60 min, 3x/day
    • Recorded mask use:
      • mask visible
      • mask used properly (over mouth & nose)
    • Estimated age, sex, activity type, distance from others
  • Phone surveys – 640 interviews
    • COVID-19 module, including mask use, attitudes, knowledge
    • Demographics, other health, employment, schooling, etc.
  • Implementation data
    • Available: number of masks given to VHTs for distribution

​

​

12

14 of 28

Analysis

  • Registered pre-analysis plan

​

  • Test whether free mask distribution affected mask use
    • Pre free mask distribution/post free mask distribution

​

  • Test whether VHT training affected mask use
    • Randomized design

​

13

15 of 28

Limitations

  • External validity
    • Study conducted in small geographic & a lull in COVID-19
    • Short follow-up period

​

  • Pre/post design mask distribution�
  • Fidelity to experimental design
    • Volunteers not observed distributing masks, not able to confirm that education intervention was delivered

​

​

​

14

16 of 28

Results�

15

17 of 28

Mask distribution data

16

  • Number of masks given to VHTs to distribute to households

​

Overall

Bumasikye

Busiu

Lukhonje

Mean

288

320

285

92

Median

250

282

250

240

Range

(136, 711)

(136, 711)

(164, 500)

(137, 561)

18 of 28

17

  • Large discrepancy between self-reports and observed mask use

​

  • Small but significant increase in mask use after distribution

​

  • Observed mask use remained low

​

19 of 28

Increased mask use during higher risk activities

18

Jakubowski et al. (2022) Preprint at CEGA Working Papers, doi:10.26085/C3XS3Q

Mask worn correctly

Mask visible

20 of 28

19

Panel A: Mask Observations

Mean in control� villages

Impact of volunteer training

95% CI

Observed having a mask

1.9

-0.1

[-1.2 – 1.1]

Observed wearing a mask

1.3

-0.3

[-1.2 – 0.6]

Panel B: Phone Surveys

 

 

 

Self-reported wearing mask to locations:

 

 

 

public place

96.5 %

1.6

[-1.2 – 4.3]

market centers

76.4 %

5.7

[-0.4 – 11.8]

religious gatherings

92.1 %

2.7

[-0.9 – 6.2]

public transportation

86.9 %

2.8

[-1.6 – 7.2]

village store

65.3 %

0.6

[-7.8 – 8.9]

Says COVID-19 spreads through:

 

 

 

surfaces

43.1 %

-0.6

[-8.9 – 7.8]

droplets

79.9 %

-4.4

[-9.7 – 0.9]

breath

50.6 %

2.7

[-6.7 – 12.0]

Beliefs about COVID-19 / masks

 

 

 

People in Uganda are at risk of infection

77.0 %

1.9

[-4.7 – 8.5]

Masks reduce spread of COVID-19

97.5 %

0.7

[-1.6 – 2.9]

COVID-19 is more serious than malaria

87.0%

3.5

[-1.8 – 8.8]

Behavior changes due to COVID

 

 

 

Number of new behaviors (range 0-14)

3.7

4.7

[-19.3 – 28.6]

Social distancing

 

 

 

Visited religious gathering outside village

34.0 %

-1.8

[-11.5 – 7.9]

Physical health

 

 

 

Experienced any COVID-19 symptoms

54.0 %

-2.5

[-11.2 – 6.2]

Mental health

 

 

 

Felt nervous, anxious, on edge

56.1 %

-3.8

[-11.2 – 3.5]

No evidence that VHT training had an impact

21 of 28

Conclusions�

20

22 of 28

Conclusion

  • Small increase in mask use following distribution
    • Suggests access to masks may have been a barrier to some
  • Mask use very low in the study region (pre & post distribution)
  • Vast discrepancy in observed vs. self-reported mask use
  • Variation in mask use by activity
    • Increased mask use in higher risk activity like commuting
  • No evidence that training volunteers improved mask recipients’ behavior, knowledge, or attitudes about masks
    • Suggests lack of information was not a barrier
  • Social norms key factor in observable health behaviors

​

21

23 of 28

22

Social desirability bias in mask use also found in sister project in Kenya

24 of 28

Health behavior change: lessons from mask studies  

  • Social desirability bias in self-reported health measures
    • Key when health behavior accepted through norms (masks, smoking)

​

  • Putting Kenya and Uganda mask study results in context
    • Large increase in masking in Bangladesh RCT Abaluck et al. Science (2022)
      • Mask distribution + education + modeling of behavior + reinforcement

​

  • What differed Uganda?
    • Context, norms
      • Baseline mean mask use: 0.5% in Uganda, 13.3% in Bangladesh
    • Lighter touch interventions in Uganda (no reinforcement or role model)

​

  • Conclusion: social norms key and need intensive interventions to affect mask use (but that is expensive)

23

25 of 28

Acknowledgements

  • Office of Prime Minister, Ministry of Health, Mbale District
  • Innovations for Policy Initiative (IPA) Uganda
  • Funding: Center for Effective Global Action via Wellspring Philanthropic Fund
  • Makerere University School of Public Health
  • University of California, Berkeley

​

​

24

26 of 28

​

Contacts:

Dr. Ronald Mulebeke

Email: mulebeker@gmail.com

Tel: +256785613467

25

27 of 28

Extra slides

26

28 of 28

Analysis

  •  

27