Cost and cost-effectiveness of newborn home visits: findings from the Newhints cluster-randomised controlled trial in rural Ghana

Summary Background Every year, 2·9 million newborn babies die worldwide. A meta-analysis of four cluster-randomised controlled trials estimated that home visits by trained community members in programme settings in Ghana and south Asia reduced neonatal mortality by 12% (95% CI 5–18). We aimed to estimate the costs and cost-effectiveness of newborn home visits in a programme setting. Methods We prospectively collected detailed cost data alongside the Newhints trial, which tested the effect of a home-visits intervention in seven districts in rural Ghana and showed a reduction of 8% (95% CI −12 to 25%) in neonatal mortality. The intervention consisted of a package of home visits to pregnant women and their babies in the first week of life by community-based surveillance volunteers. We calculated incremental cost-effectiveness ratios (ICERs) with Monte Carlo simulation and one-way sensitivity analyses and characterised uncertainty with cost-effectiveness planes and cost-effectiveness acceptability curves. We then modelled the potential cost-effectiveness for baseline neonatal mortality rates of 20–60 deaths per 1000 livebirths with use of a meta-analysis of effectiveness estimates. Findings In the 49 zones randomly allocated to receive the Newhints intervention, a mean of 407 (SD 18) community-based surveillance volunteers undertook home visits for 7848 pregnant women who gave birth to 7786 live babies in 2009. Annual economic cost of implementation was US$203 998, or $0·53 per person. In the base-case analysis, the Newhints intervention cost a mean of $10 343 (95% CI 2963 to −7674) per newborn life saved, or $352 (95% CI 104 to −268) per discounted life-year saved, and had a 72% chance of being highly cost effective with respect to Ghana's 2009 gross domestic product per person. Key determinants of cost-effectiveness were the discount rate, protective effectiveness, baseline neonatal mortality rate, and implementation costs. In the scenarios modelled with the meta-analysis results, the ICER increased from $127 per life-year saved at a neonatal mortality rate of 60 deaths per 1000 livebirths, to $379 per life-year saved at a rate of 20 deaths per 1000 livebirths. The strategy had at least a 99% probability of being highly cost effective for lower-middle-income countries in all neonatal mortality rate scenarios modelled, and at least a 95% probability of being highly cost effective for low-income countries at neonatal mortality rates of 30 or more deaths per 1000 livebirths. Interpretation Our findings show that the seemingly modest mortality reductions achieved by a newborn home-visit strategy might in fact be cost effective. In Ghana, such strategies are also likely to be affordable. Our findings support recommendations from WHO and UNICEF that low-income and middle-income countries implement newborn home visits. Funding The Bill & Melinda Gates Foundation, UK Department for International Development, WHO.


Population estimates
The 2005 census estimated the combined population of the 7 districts of the study to be 698,046 and estimated the population growth rate at 2.5%. 4

Capital
A proportion of both the capital and recurrent costs of vehicles and equipment was attributed to the intervention based on actual usage, although the full cost of purchasing the two vehicles is included in the financial cost of setup. For the project's two 4x4 vehicles, which were used in supervision of the intervention as well as in unrelated activities, vehicle log books were analysed to identify the proportion of vehicle costs attributable to the intervention. A share of equipment cost was allocated to the intervention based on the proportion of time attributed to the intervention for the individuals who used the equipment.

Staff time
To estimate the financial and economic costs of KHRC and LSHTM staff time, data on the value of salaries and benefits were combined with results of a time use study, which estimated the proportion of time each member of staff contributed to the implementation of Newhints relative to research activities or other projects. During two-week periods in November/December 2008 and again in May 2009, project staff completed daily self-reports of their time use and participated in semi-structured interviews twice a week. Following these short surveys, project staff were interviewed on their own time use and that of their colleagues to triangulate best, minimum, and maximum estimates for each staff member by 6-month period.
The sum of salaries or stipends and benefits provided was considered to reflect both the financial and economic cost of supervisors and CBSVs. The time CBSVs and their supervisors spent delivering the Newhints intervention was examined through a combination of review of records (quantitative) and indepth interviews (qualitative) with supervisors. 6 The average number of CBSV visits per period was estimated based on women's reports of how many antenatal and postnatal CBSV visits they had received. The duration of a home visit was estimated from the "direct observation sheets" which were completed by supervisors during monthly accompanied home visits with the CBSVs. The number and duration of supervision visits and accompanied home visits per month were extracted from "tally sheets", in which supervisors report on their activities. Data covered all activities from February 2009, when tally sheet collection began, until the end of the study period.

Overheads
Overheads were allocated to the intervention based on office size for utilities and based on staff numbers for management costs, with research costs excluded from each based on the results of the staff time use study.

Capital
Capital goods accounted for 14.8% of costs. Of the annualized economic costs of capital in 2009 ($30,225), 70.9% reflected vehicle costs, including two 4x4s (29.9%), fourteen motorcycles for the fourteen supervisors (37.6%) and bicycles for some (n=28) of the CBSVs (2.4%), while nearly all of the remaining annualized economic costs of capital were incurred for equipment provided to CBSVs, such as manuals and counselling cards.

Materials provided to CBSVs
All CBSVs received a package of materials which included both equipment and supplies, for which the total annualized economic cost is estimated as $19.45 (Web 100.0% 1 Quantities vary with the number of women visited per CBSV. We present an average here. 2 Very few (n=28) CBSVs received a bicycle. We present the annualized cost for a CBSV who did receive a bicycle, but do not include it the total.

Additional care-seeking
The incremental cost of care-seeking for sick newborns attributable to the Newhints intervention was $6,601, and thus represents a 3.1% increase in total costs beyond the annualized direct cost of implementing Newhints. Of the 7,786 newborns in the intervention zones, an estimated 484 more newborns in the Newhints zones were taken to health facilities than would have been in the absence of the intervention. These included 21 additional newborns for whom care was sought spontaneously by caregivers and an estimated 463 who were taken to a health facility following referral by a CBSV.
The estimate of 21 additional newborns for whom care was sought spontaneously by caregivers was derived as follows: In the baseline period (2005-7), 1.2% of caregivers in the Newhints zones reported that their baby was severely ill (280/23,221 newborns); Of these, care was sought for 52.5% (147/280 newborns). 7 In the intervention period, Newhints was estimated to have reduced the care-seeking gap by 45% (28% to 73%). 7 Multiplying these figures together gives (1.2% perceived as ill) * ((1 -52.5%) care not sought previously) * (45% care sought) * (7,786 newborns) = 20.7 additional newborns for whom care was sought beyond those for whom care would have been sought in the absence of the intervention. This estimate takes into account the higher rates of care-seeking in the Newhints zones for newborns perceived to be ill by their caregivers, but assumes that the Newhints intervention did not affect the underlying rate of true illness and caregivers' rate of recognition of severe illness. In reality, one would expect that the improved behavioural practices in the Newhints zones would lead to lower underlying true rates of illness, and that counselling during the CBSV visits would also lead caregivers to be more able to recognize danger signs, and thus be more likely to perceive illness.
The estimate of 463 newborns taken to health facilities following referral was derived as follows: Approximately 69.8% of newborns were visited by a CBSV postnatally; of these, 95.0% were assessed for danger signs; of these, 10.4% were referred; and of these, 86.0% complied with the referral. 8 As we estimate that 7,786 babies were born in the intervention zones in the 12-month study period, multiplying these figures together gives: (69.8% visited) * (95.0% assessed) * (10.4% referred) * (86.0% complied) *(7,786 newborns) = 462.9 newborns taken to a health facility following referral by a CBSV.
Amongst all babies for whom care was sought, 17.1% were admitted to hospital, 7 of whom approximately 90% were diagnosed with sepsis and approximately 10% had very low birth weight. 8 The GNHIS reimbursement rate for admitting newborns to public hospitals with sepsis in 2009 was $63.00 and with birth weight abnormalities is $68.66, resulting in a weighted average cost to the health service of $63.57 per newborn admitted, or $5,228 for the estimated 82 additional newborn admissions. At $3.42 per newborn, the cost of the estimated 401 additional newborns who were consulted as outpatients was considerably lower, at $1,374.

Decision Tree
The decision tree diagram reflects our conceptualization of the decision problem. Cost-effectiveness estimates are based on the equations presented in the main text and in the following section of the web appendix.

Equation for the incremental cost-effectiveness ratio
The calculation of the incremental cost-effectiveness ratio (ICER) per newborn life saved is given in the main text. We calculated the number of life-years saved with a 3% discount rate, no age weighting, and Ghana's life expectancy at birth in 2010 of 65 years (62.6 to 67.3) 9 as follows: Where CD: discounted total direct economic costs, CI: indirect economic costs, NIntervention: number of livebirths in the intervention area, NMRControl: NMR in the control areas, PE: protective effectiveness (risk ratio) of the intervention, d: discount rate, and L: life expectancy.

Web table 3: Model parameters and results of one-way sensitivity analysis
The table presents the parameter distribution for each variable in the probabilistic sensitivity analysis, the parameter range used for the one-way sensitivity analysis, and the results of one-way sensitivity analysis for each of the uncertain variables and key assumptions. Where parametric distributions are specified for the probabilistic sensitivity analysis (PSA), the low and high values represent the 95% confidence interval. The incremental cost-effectiveness ratio (ICER) represents the cost (in constant 2009 USD) per life year saved (LYS) with both costs and effects discounted at 3% for all analyses except the one-way analysis of the effect of the discount rate on the ICER.

Web table 4: Protective efficacy/effectiveness vs. the baseline neonatal mortality rate in four existing proof-of-principle studies and four programmatic cluster-randomized controlled trials of newborn home visits
Studies were identified and grouped in a previous systematic review and meta-analysis. 7