Table 1.

Baseline input parameters used in modeling the promotion of smartphone apps for weight loss.

Key parameter	Source and application to model	Uncertainty	Distribution and heterogeneity
Baseline population count	Statistics New Zealand (SNZ) population estimates for 2011	Nil uncertainty	Sex; age; ethnicity
All-cause mortality rates	SNZ mortality rates for 2011	Nil uncertainty	Sex; age; ethnicity
Disease-specific incidence, prevalence, case fatality rates, and remission rates	For each disease, coherent sets of incidence rates, prevalence, case fatality rates (CFR), and remission rates (zero for noncancers, the complement of the CFR for cancers to give the expected 5-year relative survival) were estimated using DISMOD II using data from New Zealand Burden of Disease Study (NZBDS), HealthTracker, and the Ministry of Health	Uncertainty: rates all ±5% SD	Log-normal; sex; age; ethnicity
Disease trends	Trends are applied to incidence, case fatality, and remission. These are switched on until 2026 and then kept constant for the remainder of the lifetimes of the modeled population	Uncertainty ±0.5% absolute change; diabetes: uncertainty ±1.5% absolute change	Normal; sex; ethnicity
Total morbidity per capita in 2011	The per capita rate of years of life lived with disability (YLD) from the NZBDS	Uncertainty ±10% SD	Log-normal; sex; age; ethnicity
Disease morbidity rate per capita	2006 NZBDS (projected to 2011); each disease was assigned a disability rate (DR; by sex and age) equal to YLDs for that disease (scaled down to adjust for comorbidities) from the 2006 NZBDS projected forward to 2011, divided by the disease prevalence. This DR was assigned to the proportion of the cohort in each disease state	Uncertainty: ±10% SD	Normal; sex; age
Health system costs	Linked health data (hospitalizations, inpatient procedures, outpatients, pharmaceuticals, laboratories, and expected primary care usage) for each individual in New Zealand for the period 2006 to 2010 had unit costs assigned to each event, and then health system costs (NZ $2011) were estimated	Estimated at SD ±10% of the point estimate	Gamma; sex; age
Time lags for intervention effect	It takes time for a change in body mass index (BMI) to impact on disease incidence. As there are no precise data on just how long these are, we have used wide windows of time lags. For cancers, the time lag is assumed to range between 10 and 30 years. For CHD, stroke, diabetes, and osteoarthritis (the noncancers), the time lag is assumed to be shorter and ranges between 0 and 5 years. Wide uncertainty is included around these estimates	Uncertainty: ±20% SD	Normal
BMI theoretical minimum risk exposure level (TMREL)	TMREL is the level of risk exposure that is theoretically possible and minimizes overall risk and is derived from the latest Global Burden of Disease 2013 study [10]. This allows us to estimate how much of the disease burden could be lowered by shifting the distribution of a risk factor to the level that would lead to the greatest improvement in population health	Uncertainty: uniform distribution between 0 and 1	Uniform
Height of the New Zealand adult population (for BMI calculations)	Mean and SD of height from the New Zealand Adult Nutrition Survey 2008 to 2009 [11]	Uncertainty using reported SD	Normal; sex; ethnicity