Publications
Publications
The Health Effects of Motorization
Authors: Bhalla, K.
Journal: International Journal of Injury Control and Safety Promotion, 2011.
Download Article: Click Here
Non-traditional data sources for injury control: an agenda for action in Ghana, Injury Prevention, 2012
Authors: Bhalla, K., Koranteng, A., Mock, C., Afukaar, F., Appiah, N., and Ebel, B.
Journal: Injury Prevention, 2012
Download Article: Click Here
The global injury mortality data collection of the GBD-Injury Expert Group: a publicly accessible research tool
Authors: Bhalla, K., Harrison, J., Fingerhut, L., Shahraz, S., Abraham, J., and Yeh, P-H, on behalf of the Global Burden of Disease Injury Expert Group
Journal: International Journal of Injury Control and Safety Promotion, 2011.
Download Article: Click Here
An assessment of the availability and quality of cause of death data for estimating the global burden of injuries
Authors: Bhalla, K., Harrison, J., Shahraz, S., Fingerhut, L., on behalf of the Global Burden of Disease Injury Expert Group
Journal: Bulletin of the World Health Organization, 2011
Download Article: Click Here
Incidence of road injuries in Mexico: country report
Authors: Bartels, D., Bhalla, K., Shahraz, S., Abraham, J., Lozano, R., and Murray, C.J.L.
Journal: International Journal of Injury Control and Safety Promotion, 2010.
Download Article: Click Here
Methods for developing country level estimates of the incidence of deaths and non-fatal injuries from road traffic crashes
Authors: Kavi Bhalla, Saeid Shahraz, David Bartels and Jerry Abraham
Journal: International Journal of Injury Control and Safety Promotion, 2009, 16(4): 239-248.
Download Article: Click Here
Data sources for improving estimates of the global burden of injuries: call for contributors
Authors: Kavi Bhalla, James Harrison, Jerry Abraham, Nagesh Borse, Ronan Lyons, Soufiane Boufous, Limor Aharonson-Daniel, on behalf of the Global Burden of Disease Injury Expert Group
Journal: PLoS Medicine, January 2009, 6(1).
Download Article: Click Here
Building national estimates of the burden of road traffic injuries in developing countries from all available data sources: Iran
Authors: Bhalla, K., Naghavi, M., Shahraz, S., Bartels, D., and Murray, C.J.L.
Journal: Injury Prevention, 15: 146-149.
Download Article: Click Here
Adverse health outcomes of road traffic injuries in Iran after rapid motorization,
Authors:Naghavi M., Shahraz S., Bhalla K., Jafari N., Pourmalek F., Bartels D, Puthenpurakal J.A., and Motlagh M.E.
Journal: Archives of Iranian Medicine, May 2009;12(3):284-94.
Estimating the distribution of external causes in hospital data from injury diagnosis
Authors: Bhalla, K., Shahraz, S., Naghavi, M., Lozano, R., and Murray, C.
Journal: Accident Analysis and Prevention, 2008, 20: 1822-1829.
Download Article: Click Here
A new initiative for monitoring road traffic injury metrics in developing countries (Poster)
Authors: Kavi Bhalla, Saeid Shahraz, Mohsen Naghavi, Marc Shotten, Tony Bliss and Chris Murray
Presented at: 9th World Conference on Injury Prevention and Safety Promotion, Merida, Mexico, March 2008.
Abstract: Introduction: Reliable statistics on road traffic injuries (RTI) are an essential input for describing the public health burden of injuries, evaluating the impact of safety policies, and benchmarking achievements. While injury surveillance systems are common in high income countries, most low and middle income countries are unlikely to have such capacity for several decades. In the interim, we (the Harvard Initiative for Global Health in partnership with the World Bank’s Global Road Safety Facility) are developing a publicly available cross-national RTI database by harmonizing injury statistics from a wide array of data sources.
Material and Methods: We are working with the injury community to develop a publicly available database containing best estimates of national level road traffic deaths and injuries broken down by age, gender, location (e.g. urban, rural, type of road), victim type (e.g. pedestrian, motorcycle, occupant) and vehicle type. These estimates will be based on analyses of all relevant data sources in 18 regionally representative countries. These data sources include crime reports, hospital records, crematorium records, insurance claims, and health and economic surveys. We are developing methodologies that improve the accuracy of these datasets by accounting for known biases, make the various data sources compatible, and allow national estimates based on sub-national data.
Results and Conclusions: We present the proposed framework with sample results from six developing countries. We call on injury researchers and policy makers to collaborate with us in our mission to develop this global public good.
Comparing the burden of road traffic injuries in two developing countries: Iran and Mexico (Presentation)
Authors: Mohsen Naghavi, Saeid Shahraz, Kavi Bhalla, Rafael Lozano, and Chris Murray
Presented at: 9th World Conference on Injury Prevention and Safety Promotion, Merida, Mexico, March 2008.
Abstract: We compare and contrast the burden of road traffic injuries in two developing countries, Iran and Mexico, using a standardized injury metrics framework.
Materials and Methodology: We analyze the following in each country: death registration, hospital discharge, emergency room visits, police reports, and health surveys. We estimate the number of deaths, outpatient visits, inpatient visits, and no-care cases due to road traffic injuries in the two countries. We compare these results separately by age, sex, and victim type and compute the burden of RTI by estimating years of life lost to mortality and years lost to disability based on methods developed in the Global Burden of Disease and Injury studies.
Results: Although Mexico and Iran have many common characteristics (such as income per capita, population age structure, all-cause death rate) and are typical of countries in health transition, they have vastly differing RTI metrics. For instance, RTI death rates in Iran (44 per 100,000 people) are more than twice the rate in Mexico (17). While car occupants and motorcycle riders are the leading victims in Iran (29.2% and 28.6%, respectively), in Mexico, pedestrians make up the bulk of RTI victims (54%).
Conclusions: Our analysis reveal vast differences between Iran and Mexico emphasizing the need for in-depth country level analysis for understanding causal mechanisms, identifying risk factors and designing safety strategies.
Improving the quality of injury statistics by using regression models to redistribute ill-coded events (Poster)
Authors: Saeid Shahraz, Kavi Bhalla, Mohsen Naghavi, Rafael Lozano, and Chris Murray
Presented at: 9th World Conference on Injury Prevention and Safety Promotion, Merida, Mexico, March 2008.
Abstract: Cause of death data sources in low and middle income countries often have a large number of cases coded to ill-defined causes. Proportional redistribution of unknown events among other categories ignores the possibility that the set of ill-defined cases may be biased. We compare methods that correct these biases by using the additional information that may be available about the victim in case level data (individual records).
Method: We compare three strategies for redistribution of ill-defined events: proportional redistribution, multiple-imputation (using the software package Amelia), and multinomial-logistic regression model. The multiple-imputation and multinomial regression models rely on the use of the additional variables (such as age, sex, province, insurance type, education level, time and place of death) available in the vital registration dataset. We evaluate the performance of these methods by application to a validation dataset of a random sample of known cases from death registration records from Mexico (2004).
Results: Multinomial regression models perform better than multiple-imputation and proportional redistribution. However the relative difference in performance between the three methods is small. Multiple imputations have a tendency to over-assign cases to events with a small number of cases.
Conclusions: When case level data is available, regression models that take advantage of other covariates in the dataset should be used to redistribute ill-defined events over known cause groups. However, when only tabulations are available, proportional redistribution of unknowns provides reasonably accurate results.
Estimating distribution of external causes of injury from datasets that only record nature of injury (Presentation)
Authors: Saeid Shahraz, Kavi Bhalla, Mohsen Naghavi, Rafael Lozano, and Chris Murray
Presented at: 9th World Conference on Injury Prevention and Safety Promotion, Merida, Mexico, March 2008.
Abstract: In most developing countries, hospital records contain information about injuries sustained by victims (e.g. fractures, burns, etc) but not the external causes of injury (e.g., falls, road crashes, etc). We develop a method for estimating the number of hospital admissions due to each external cause based on the nature of injuries.
Materials and Methods: We evaluate three strategies for mapping injuries to external causes by comparing their performance on a validation dataset in which both injuries and external causes are known. The validation dataset consists of a training dataset of 50,000 cases and several application datasets of 1000 cases each, randomly sampled from the 2004 Mexico Ministry of Health hospital discharge dataset.
Mapping Strategies: We proportionately redistribute unknown cases based on the age and sex distribution in the training dataset. We fit a multinomial logistic regression model that estimates external cause based on nature of injury, age, sex, province, insurance type, etc, using the training dataset and apply to application dataset. Bayesian Updating: We update the prior probability distribution of external causes based on proportional redistribution in the training dataset by using the information available about injuries. Thus, for e.g., p(fall/head_injury)=p(fall)*p(head_injury/fall)/p(head_injury)
Results: The multinomial regression models as well as Bayesian updating are significant improvements over proportional redistribution. Although the regression model performed marginally better than Bayesian updating, it was computationally intensive and not feasible for large datasets.
Conclusions: Bayesian updating is a computationally effective method of estimating the number of injuries due to external causes when such information is missing or unavailable, as is often the case with hospital records.
A Risk-Based Method for Modeling Traffic Fatalities
Journal: Risk Analysis, 27(1), 2007
Authors: Kavi Bhalla, Majid Ezzati, Ajay Mahal, Joshua Salomon, and Michael Reich
Download Article: Click Here
Road Traffic Injuries - the growing gap (Presentation)
Presented at: Second United Nations Global Road Safety Week Stakeholders Forum, Geneva, Switzerland, April 2007
Author: Kavi Bhalla
Estimating the potential impact of safety policies on road traffic death rates in developing countries (Poster)
Authors: Kavi Bhalla, Saeid Shahraz, Mohsen Naghavi and Christopher Murray
Presented at: 9th World Conference on Injury Prevention and Safety Promotion, Merida, Mexico, March 2008.
Abstract: We use econometric models to separate the effects of income growth and safety policies in historic data from rich countries and use the results to predict future road traffic injury (RTI) death rates in poor countries with and without safety policies.
Methodology: We obtained cause of death data for 18 high income OECD countries from the WHO mortality database and constructed time-series-cross-sections (1955-2002) of RTI death rates for 16 age-sex groups. For each, we develop time- and country- fixed effects regression models of RTI death rates controlling for income effects, urbanization and population density. We use the method of panel corrected standard errors developed by Beck and Katz and remove serial autocorrelation by including a lagged dependent variable. Using the estimated coefficients of time, which are a measure of policy impact in rich countries, we use the model to project the future (2002-2050) of poor countries under two scenarios: with and without these policy impacts.
Results: Our models show statistically significant declines in RTI death rates in rich countries beginning in 1970 even after controlling for income growth. If these policy effects are ignored, our projections until 2050 show steadily increasing RTI deaths in poor countries. However, when we apply post-1970 time trends from rich countries to the future of poor countries, the rise in RTI deaths is reversed.
Conclusions: Our analysis clearly identifies a turning point in the RTI history of rich countries which we attribute to the adoption of safety policies.