Improved health outcomes in urban slums through infrastructure upgrading

Three CBAs (Butala 2010; Galiani 2007; Soares 2005) had mixed risk of selection bias, as they were not randomised but employed difference-in-differences and adjusted for potential confounders.

“Three CBAs (Butala 2010; Galiani 2007; Soares 2005) were graded as having a mixed risk of selection bias, as they were not randomised but employed a DiD approach to minimise selection bias and adjusted for potential confounders.”

Three main studies (Butala 2010; Galiani 2007; Taylor 1987) and three supporting studies (Aiga 2002; Joshi 2002; Parikh in press) reported community involvement, either through slum residents paying for intervention delivery or fees for resulting infrastructure such as water supply.

“Three main studies (Butala 2010; Galiani 2007; Taylor 1987) and three supporting studies (Aiga 2002; Joshi 2002; Parikh in press) reported community involvement, either in the form of slum residents paying for the intervention delivery or the fees for the resulting infrastructure such as water supply (see Figure 3 and Figure 4).”

Five main studies were identified: one cluster-RCT (Gonzalez-Navarro 2010) and four cluster-CBAs (Butala 2010; Galiani 2007; Soares 2005; Taylor 1987) using difference-in-differences analyses and adjustment for confounders.

“Five main studies were identified for inclusion: one cluster‐RCT (Gonzalez‐Navarro 2010) and four cluster‐CBAs (Butala 2010; Galiani 2007; Soares 2005; Taylor 1987) using difference in difference analyses and adjustment for confounders.”

Partnerships between different combinations of government, civil society, private industry, NGOs, and charitable trusts were described in both main and supporting studies (Butala 2010; Galiani 2007; Joshi 2002; Parikh in press).

“Partnerships between different combinations of government, civil society, private industry, NGOs and charitable trusts were also described in both the main and supporting studies (Butala 2010; Galiani 2007; Joshi 2002; Parikh in press).”

Among the three CBAs, outcomes differed: Butala 2010 examined waterborne illnesses, Taylor 1987 examined mean household income, and Soares 2005 measured mortality (infant, homicide, sanitation-related), income, and illiteracy.

“Outcomes differed between studies. Butala 2010 examined effects on waterborne illnesses, whereas Taylor 1987 examined effect on mean household income and Soares 2005 measured mortality (infant, homicide and sanitation‐related), income and illiteracy.”

Three CBAs compared broad slum upgrading strategies involving physical upgrading and wider living environment strategies to no intervention (Butala 2010; Soares 2005; Taylor 1987).

“In brief, three CBAs compared the impact of broad slum upgrading strategies, involving physical upgrading and wider living environment strategies, to no intervention (Butala 2010; Soares 2005; Taylor 1987).”

Follow-up after intervention completion was unclear in Butala 2010 and Soares 2005.

“Follow‐up after intervention completion was unclear in the two remaining main studies (Butala 2010; Soares 2005).”

Four main studies (Butala 2010; Galiani 2007; Gonzalez-Navarro 2010; Soares 2005) allocated the intervention to slum communities (clusters) rather than individual slum dwellers and used appropriate analyses.

“Four main studies allocated the intervention to slum communities (that is clusters) rather than individual slum dwellers and used appropriate analyses (Butala 2010; Galiani 2007; Gonzalez‐Navarro 2010; Soares 2005).”

Three main studies (Butala 2010; Galiani 2007; Taylor 1987) did not report information on fidelity.

“Of the main studies, three did not report information on fidelity (Butala 2010; Galiani 2007; Taylor 1987).”

In Butala 2010, the slum upgrading intervention resulted in a significant decrease in waterborne illness claims as opposed to non-waterborne illness claims in the conservative model (RR 0.64, 95% CI 0.41 to 0.98; P=0.04).

“The estimated relative risk ratio (RR) indicated that the slum upgrading intervention resulted in a significant decrease in waterborne illness claims as opposed to non‐waterborne illness claims for both the broad model (RR 0.62; 95% CI 0.61 to 0.64; P = 0.04) and conservative model (RR 0.64; 95% CI 0.41 to 0.98; P = 0.04) (Butala 2010).”

In Butala 2010, the slum upgrading intervention resulted in a significant decrease in waterborne illness claims as opposed to non-waterborne illness claims in the broad model (RR 0.62, 95% CI 0.61 to 0.64; P=0.04).

“The estimated relative risk ratio (RR) indicated that the slum upgrading intervention resulted in a significant decrease in waterborne illness claims as opposed to non‐waterborne illness claims for both the broad model (RR 0.62; 95% CI 0.61 to 0.64; P = 0.04) and conservative model (RR 0.64; 95% CI 0.41 to 0.98; P = 0.04) (Butala 2010).”

Findings from Butala 2010 may not be representative of the most disadvantaged slum residents, who may not be able to afford health insurance.

“findings may not be representative of the most disadvantaged slum residents, who may not be able to afford health insurance.”

The outcome measure in Butala 2010 was considered unreliable because assumptions about changes in the proportion of waterborne illnesses rely on no significant change in non-waterborne illnesses.

“This outcome measure was considered unreliable because assumptions about the changes in the proportion of waterborne illnesses relies on no significant change in non‐waterborne illnesses.”

The criteria used by researchers to code illness claims in Butala 2010 were not reported.

“The criteria used by researchers to code illness claims were not reported, and findings may not be representative of the most disadvantaged slum residents, who may not be able to afford health insurance.”

Butala 2010 was the only study that used health insurance claims to measure health, expressed as a ratio of waterborne to non-waterborne diseases.

“Butala 2010, a CBA, was the only study that used health insurance claims to measure health, expressed as a ratio of waterborne compared to non‐waterborne diseases.”

A range of terminologies were used to describe the implemented projects.

“A range of terminologies were used to describe the implemented projects, including slum upgrading (Butala 2010), slum improvement (Abelson 1996; Moitra 1987), Kampung Improvement Programme (KIP – Milone 1993; Taylor 1987), Zonal Improvement Programme (ZIP, Aiga 2002), Barrio Escopa pilot project (De Leon 1986), Small Town Urban Development Project (STUDP ‐ Milone 1993), and Favela‐Bairro (Soares 2”

Micro‐credit loans were included within the programme of one main study.

“Regarding co‐interventions, micro‐credit loans were included within the programme of one main study (Butala 2010).”

Only one main study described anecdotal evidence suggesting that a majority of ancillary health interventions did not persist after construction of the basic infrastructure as per the slum upgrading programme.

“Only one main study (Butala 2010), a CBA, described anecdotal evidence suggesting that a majority of ancillary health interventions did not persist after construction of the basic infrastructure as per the slum upgrading programme.”

Broad slum upgrading strategies involving multicomponent physical environment and infrastructure interventions plus other interventions within the living environment were investigated in two of the five main studies and four supporting studies.

“Broad slum upgrading strategies involving multicomponent physical environment and infrastructure interventions plus other interventions within the living environment (such as health and social service provision or health behaviour, education) were investigated in two of the five main studies (Butala 2010; Taylor 1987) and four supporting studies (Abelson 1996; De Leon 1986; Joshi 2002; Moitra 1987”

Health education co‐components were included in the slum upgrading programmes of two main studies and secure tenure in one study.

“Health education co‐components were included in the slum upgrading programmes of two main studies (Butala 2010; Taylor 1987) and secure tenure in one study (Butala 2010).”

To reduce the risk of selection bias such limitations were likely to introduce, the majority of main studies utilised difference in difference analyses, propensity score matching and multivariate analyses that controlled for confounders.

“To reduce the risk of selection bias such limitations were likely to introduce, the majority of main studies utilised difference in difference analyses (Butala 2010; Galiani 2007; Gonzalez‐Navarro 2010; Soares 2005), propensity score matching (Soares 2005) and multivariate analyses that controlled for confounders (Galiani 2007; Gonzalez‐Navarro 2010; Soares 2005).”

Main studies were conducted in Asian countries including India and Indonesia, and Latin America including Argentina, Mexico and Brazil.

“Main studies were conducted in Asian countries including India (Butala 2010) and Indonesia (Taylor 1987), and Latin America including Argentina Galiani 2007), Mexico (Gonzalez‐Navarro 2010) and Brazil (Soares 2005).”

While the majority of main studies followed up outcomes a number of years after baseline measures, two studies did not clarify when the slum upgrading intervention was completed, making it difficult to judge the length of follow‐up post‐intervention delivery.

“While the majority of main studies followed up outcomes a number of years after baseline measures, two studies did not clarify when the slum upgrading intervention was completed, making it difficult to judge the length of follow‐up post‐intervention delivery (Butala 2010; Soares 2005).”

Of the main studies, one RCT was considered to have high external validity, and three CBAs mixed generalisability.

“Of the main studies, one RCT was considered to have high external validity (Gonzalez‐Navarro 2010), and three CBAs mixed generalisability (Butala 2010; Galiani 2007; Soares 2005).”