Presentation is loading. Please wait.

Presentation is loading. Please wait.

Drivers of Unit Cost Variation in Voluntary Medical Male Circumcision in Sub-Saharan Africa: A meta-regression analysis Drew Cameron UC Berkeley IAEN.

Similar presentations


Presentation on theme: "Drivers of Unit Cost Variation in Voluntary Medical Male Circumcision in Sub-Saharan Africa: A meta-regression analysis Drew Cameron UC Berkeley IAEN."— Presentation transcript:

1 Drivers of Unit Cost Variation in Voluntary Medical Male Circumcision in Sub-Saharan Africa: A meta-regression analysis Drew Cameron UC Berkeley IAEN pre-conference Amsterdam, Netherlands 20 July 2018

2 Overview Goal: Use published VMMC costing data to create a predictive model to extrapolate to settings where no data exists VMMC = Proof of Concept Search and Screening Extraction process Data preparation Exploration & analysis Limitations & next steps

3 PRISMA for published HIV studies
Web of Science 2,682 Cochrane 810 PubMed 6,288 Embase 10,950 NHS EED 466 LILACS 163 Gray Lit & Google Scholar 2,399 Database Searching Total 21,539 Records identified 29 Records from other sources (Avenir, UCR, LSHTM, DCP3, snowball) 471 Unique Records (de-duped) 11,717 The following searches, validation, and snowball sampling activities were run between January , and October 2, 2016. Records Excluded (non relevant topics, low yield hits) 9,199 HIV Articles Assessed for Eligibility 2,566 VMMC Studies Extracted 58 VMMC Studies Finally Included (modeled & duplicate data removed) 29

4 Data Extracted 58+ MC studies extracted (2016-7)
VMMC Studies Extracted 58 VMMC Studies Included (modeled & duplicate data removed) 29 58+ MC studies extracted (2016-7) Early analysis & lessons learned Validated by analyst Quality assurance (early 2018) Removed 29 studies (29 total) Costs modeled or duplicative Insufficient reporting (gray lit)

5 Unit of observation Average unit cost of VMMC procedure (per patient)
Country Studies Unit costs No. sites Botswana 1 2 Kenya 6 251 Lesotho 4 Mozambique Namibia 7 49 Rwanda +2 South Africa 11 31 Swaziland Tanzania 14 Uganda 5 10 65 Zambia 23 25 Zimbabwe 13 Total 29 89 503 Average unit cost of VMMC procedure (per patient) often average of multiple facilities per unit cost Weight by number of facilities per cost 2016 USD (WB GDP price deflator; exchange rates)

6 Geographic distribution of unit costs
Figure 2. Heatmap of mean unit cost by country Country Mean unit cost Botswana $125.15 Kenya $44.79 Lesotho $64.50 Mozambique $32.31 Namibia $56.73 Rwanda $43.96 South Africa $107.68 Swaziland $61.79 Tanzania $69.25 Uganda $34.75 Zambia $70.06 Zimbabwe $61.71

7 Geographic distribution of unit costs
Figure 2. Heatmap of mean unit cost by country Country Mean unit cost Botswana $125.15 Kenya $44.79 Lesotho $64.50 Mozambique $32.31 Namibia $56.73 Rwanda $43.96 South Africa $107.68 Swaziland $61.79 Tanzania $69.25 Uganda $34.75 Zambia $70.06 Zimbabwe $61.71

8 Data preparation Removed unit costs with poor reporting (n=48)
No input costs reported (n=38) Personnel (direct service delivery; support) & traded goods (i.e. kits) not reported (n=10) 41 total unit costs remain with sufficient data

9 Unit cost variation by country

10 Analysis goals Review: Use published unit cost estimates to predict VMMC costs in countries/settings/years with no evidence Using mean and quantile regression Use variables that are good predictors of unit cost variation (extract: urbanicity, facility type, ownership) Additional data Per capita GDP by country-year (World Bank) Annual PEPFAR / Global Fund funding by country-year (IHME) MC coverage by country (Morris et al. 2016) HIV prevalence; ART coverage by country-year (World Bank)

11 Covariate missingness
Facility Type Freq. Percent Clinic 3 7.3 Mixed (clinics & hospitals) 7 17.1 Hospitals 23 56.1 Mobile / community based 8 19.5 Ownership Freq. Percent Public 21 51.2 Mixed (public & private) 5 12.2 Private / NGO 15 36.6 Urbanicity Freq. Percent Urban / peri-urban 22 53.7 Mixed (urban & rural) 9 22.0 Rural 10 24.4

12 Covariate missingness
Facility Type Freq. Percent Clinic 3 7.3 Mixed (clinics & hospitals) 7 17.1 Hospitals 23 56.1 Mobile / community based 8 19.5 Ownership Freq. Percent Public 21 51.2 Mixed (public & private) 5 12.2 Private / NGO 15 36.6 Urbanicity Freq. Percent Urban / peri-urban 22 53.7 Mixed (urban & rural) 9 22.0 Rural 10 24.4

13 Focus on cost inputs Focused on predictive models of key inputs (personnel, supplies, residuals) Start with predicting personnel costs

14 Model Selection Attempted multiple models, functional forms, specification tests, robustness checks OLS (log – log) Country fixed effects GLM Random & mixed effects models Quantile regression

15 GLM (link log; family gamma)
VARIABLE (1) (2) (3) (4) (5) (6) (7) Key info Year of data collection (2016 reference) -0.397 -1.003 0.264 -3.317*** -3.047*** -3.743*** -4.402*** GDP per capita (mean centered) 0.011 0.012** 0.008*** 0.009*** 0.000 -0.002 GDP per capita^2 (mean centered) -0.000 -0.000** -0.000*** 0.000** Urbanicity (urban omitted) Mixed -8.912*** 5.456*** 6.817** 5.605*** 6.513*** Rural 10.440* 5.107 2.839 2.483 2.531 Ownership (public omitted) *** *** *** *** Private / NGO -0.201 -0.061 19.981 14.648 Facility type (clinic omitted) 3.807 1.450 Hospital 2.168 -1.261 Mobile / community based 4.794 1.263 Health system HIV prevalence of pop (age 15-49) *** 77.945 MC coverage of men (all ages) -7.680 4.391 ART coverage of HIV+ pop (age 15-49) 31.928** 36.548** Donor engagement PEPFAR funding (per capita) -0.467 Global Fund funding (per capita) 0.949 BIC Observations 41 Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1

16 GLM (link log; family gamma)
VARIABLE (1) (2) (3) (4) (5) (6) (7) Key info Year of data collection (2016 reference) -0.397 -1.003 0.264 -3.317*** -3.047*** -3.743*** -4.402*** GDP per capita (mean centered) 0.011 0.012** 0.008*** 0.009*** 0.000 -0.002 GDP per capita^2 (mean centered) -0.000 -0.000** -0.000*** 0.000** Urbanicity (urban omitted) Mixed -8.912*** 5.456*** 6.817** 5.605*** 6.513*** Rural 10.440* 5.107 2.839 2.483 2.531 Ownership (public omitted) *** *** *** *** Private / NGO -0.201 -0.061 19.981 14.648 Facility type (clinic omitted) 3.807 1.450 Hospital 2.168 -1.261 Mobile / community based 4.794 1.263 Health system HIV prevalence of pop (age 15-49) *** 77.945 MC coverage of men (all ages) -7.680 4.391 ART coverage of HIV+ pop (age 15-49) 31.928** 36.548** Donor engagement PEPFAR funding (per capita) -0.467 Global Fund funding (per capita) 0.949 BIC Observations 41 Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1

17 Conclusions Pro Good scope of countries included
Consistent relationships with GDP Per Capita Year Con VMMC = small sample size for analysis (n=41) Not enough good reporting of other characteristics Mean data masks variation in unit costs within countries/studies Extracted MC data alone insufficient for prediction


Download ppt "Drivers of Unit Cost Variation in Voluntary Medical Male Circumcision in Sub-Saharan Africa: A meta-regression analysis Drew Cameron UC Berkeley IAEN."

Similar presentations


Ads by Google