Principles and applications of forensic epidemiology ...

Law, Probability and Risk Advance Access published September 10, 2015 Law, Probability and Risk (2015) 0, 1–10

doi:10.1093/lpr/mgv010

Principles and applications of forensic epidemiology in the medicolegal setting MICHAEL D. FREEMANy Departments of Public Health & Preventive Medicine and Psychiatry, Oregon Health & Science University School of Medicine, Portland, Oregon, USA and Department of Forensic Medicine, HEALTH, Aarhus University, Aarhus, Denmark AND

The discipline of forensic epidemiology, a branch of forensic medicine, provides a systematic approach to the assessment of general and specific (individual) causation, with the results suitable for presentation in a court of law. In the present paper some of the methods utilized in forensic epidemiology are described, along with examples of how such methods can be reliably applied to the evaluation of specific causality in criminal and civil matters. Included in the discussion is the presentation of two case studies in applied forensic epidemiology; one in a civil action for medical negligence, and the other in a homicide investigation. Keywords: forensic epidemiology, Hill criteria, comparative risk, specific causation.

1. Introduction It is widely recognized that unreliable evidence is a significant problem in the forensic sciences generally, and in forensic medicine specifically (Colville-Ebeling, 2014; National Research Council, 2009; Pollanen, 2012). One of the explanations for this phenomenon is the lack of widely accepted standards and methods for common tasks performed in a forensic setting. Determination of the cause of injury or disease is a pivotal issue in virtually all criminal and civil actions, and one that is often vigorously contested. Despite this fact, there is a lack of widely adopted standards regarding what constitutes scientifically valid evidence of causation, as well as a means of quantifying and weighing evidence of causation. The single largest explanation for this state of affairs is the fact that causation cannot be observed, and thus conclusions of causation are not observations but rather inferences based on a presumed degree of association between an exposure and an injury (Holland, 1986). The lack of a generally accepted systematic approach to what is essentially an exercise in probabilistic reasoning results in the reliance by lay fact finders (i.e. judge and jury) on what is potentially speculative and unreliable evidence regarding causation. Outside of a forensic or legal setting, causal evaluations are most commonly performed in a medical setting by physicians. This is because the determination of the diagnosis of the condition for which the cause is sought is the responsibility of the physician, not because clinicians are trained in causal y

Corresponding author: [email protected] ! The Author [2015]. Published by Oxford University Press. All rights reserved

Downloaded from http://lpr.oxfordjournals.org/ by guest on September 11, 2015

MAURICE ZEEGERS Department of Complex Genetics, NUTRIM School of Nutrition and Translational Research in Metabolism, Maastricht University Medical Centre+, Maastricht, The Netherlands and the Maastricht Forensic Institute, The Netherlands

2 of 10

M. D. FREEMAN AND M. ZEEGERS


methodology (O’Neal, 2001). An exception is seen in the practice of forensic pathology, where the primary purpose of the post-mortem examination is to determine the manner and cause of death. In this setting, when there is a high degree of association between the diagnosis and the cause of the death (a gunshot wound to the head, for example), the determination of causation is easily made as a matter of common sense. This is because the high degree of association of the causal relationship tends to rule out competing causes. In the example of a gunshot wound to the head, causation is obvious because such injuries are nearly always fatal, and the probability of an alternative cause of death coinciding with the time of the gunshot wound is exceedingly low in most circumstances. In contrast, the cause of death in a patient with pneumonia, an 80% occlusion of the left coronary artery, and who received an intravenous injection of a narcotic 30 min before going into respiratory arrest, cannot and should not be determined as a matter of common sense. In such a circumstance the only causal analysis that can yield valid and repeatable results is the assessment and comparison of the risk of death associated with each of the plausible causes. This form of causal analysis is necessitated by a relatively low degree of association between the death and the cause. Risk is a population-based metric, defined as the probability or chance that an event will occur in the future. The field of study from which risk is estimated is epidemiology. Epidemiology is broadly described as the branch of medicine dedicated to the study of the cause of disease and injury in populations. Epidemiologic study examines the relationships between exposures and outcomes (and vice versa), and describes the results in terms of frequencies, rates, and probabilities. Epidemiologists use standardized methods to describe disease and injury occurrence in specified populations in order to identify populations that are at higher risk than others, and to evaluate factors that may account for the risk differences. In assessing causes, epidemiologists consider components of cause both individually and collectively, as well as which components are necessary (required) for causation, and the components that are sufficient for causation (Rothman and Greenland, 2005). Although a primary function of epidemiology is to understand the causes of disease and injury in populations, epidemiology is largely silent about methods for investigating the cause of disease and injury in individuals (specific causation). Despite this fact, when there is a low degree of association between an injury observed in an individual and a suspected cause, causal assessment requires the quantification and comparison of risks acting on the individual at the time of the injury. Herein lies a gap between the type of information often needed to assess causality (risk), and the ability to access and process this information by the experts who most often provide expert opinions on causation in a forensic setting (physicians). This gap is often filled with opinion that is lacking in a factual or reliable basis. An illustration of the problems created by this gap can be seen in the previous example of the hospitalized patient with pneumonia and an 80% coronary artery blockage, who died within 30 min of the injection of a narcotic analgesic. An autopsy might list the cause of death as pneumonia, with contributing causes given as coronary artery disease and possible allergic reaction to a narcotic. Neither the diagnoses (pneumonia and moderate coronary artery disease) nor the exposure (therapeutic level of narcotic medication) are highly associated with death (i.e. the conditions and medication rarely result in death when they are present), and thus the pathologist has no evidence regarding which of the possible causes is the most probable cause of death (>50% contribution). In the event that a question arose regarding the degree of contribution of the narcotic to the cause of the patient’s death it would remain unanswered. If, however, it was known from reliable epidemiologic studies that the risk of death from pneumonia was 1 in 100, the risk of death from the coronary artery disease was 1 in 1000, and the risk of death from the narcotic was 1 in 20 000, then the contribution of

PRINCIPLES AND APPLICATIONS OF FORENSIC EPIDEMIOLOGY

3 of 10

2. Principles of a specific causation analysis Broadly speaking, there are two steps to a specific causal evaluation; the first step is to determine if the association between the exposure and the outcome of interest is plausibly causal. Realistically, most causal evaluations that might involve an FE analysis concern relationships that are plausible. This first step of this process, also considered to be the general causal relationship in legal settings, is not specific to the individual; it only addresses the question ‘could the exposure have caused the injury?’ If the plausibility question can be answered affirmatively, then the analysis can progress to the second step, which is specific to the individual. This step is intended to answer the question ‘what is the probability of that exposure having caused this injury in this individual?’ This part of the analysis incorporates predictive and relevant information about the nature and intensity of the exposure, the proximity of the temporal or spatial association between the exposure and the injury, and the medical and historical characteristics of the individual. As important as it is to quantify the risk of injury associated with the exposure, it is just as important to be able to answer the corollary question of ‘what is the probability that the individual would have the injury at the same time if the exposure had not occurred?’ The results of an injury causation analysis can be quantified in terms of a conditional probability ratio, also known as a comparative risk ratio (CRR), as follows: CRR ¼

pðO jEÞ pðO j E¯ Þ

Where p ¼ probability; O ¼ injury outcome of interest; E ¼ exposure to the hazard of interest; and E¯ ¼ non-exposure to the hazard of interest The vertical line indicates the condition of the equation, such that it reads ‘the probability of the injury outcome given the exposure versus the probability of the injury outcome given no exposure’. In order to consider a CRR as an indication of a causal relationship that is not due to random effects, the estimate should be accompanied by a 95% confidence interval that does not include 1.0 in its lower


the narcotic exposure to the patient’s death, relative to competing causes of death, could be quantified and presented to a fact finder. The discipline of forensic epidemiology (FE), a branch of forensic medicine, is directed at filling the gap between clinical judgment and epidemiologic data and methods in the evaluation of both general and specific causation in civil and criminal matters (Koehler and Freeman, 2014; Freeman et al., 2008). The purpose of an FE causal analysis is to provide an evidence-based foundation for an opinion regarding comparative or relative risk of specific causation, suitable for presentation in a medicolegal setting. The assessment of causality requires knowledge of multiple disciplines in addition to epidemiology; including pathology and pathophysiology, toxicology and pharmacology, inter alia. The objective of the present discussion is to describe the systematic and evidence-based approach to the evaluation of specific causation utilizing FE methods, and to provide several examples describing how such an investigation fits within the general framework of forensic medicine, including two case study presentations.

4 of 10


bound. This allows for the conclusion that the point estimate is ‘statistically significant’ at the p 4 0.05 level as the standard. The CRR can be converted to an Attributable Probability under the Exposed (APe) or Probability of Causation (PC), as follows: ðCRR $ 1Þ % 100% ¼ APe &PC CRR

2.1 The Hill viewpoints (criteria) In some investigations of specific causation, the plausibility of the relationship is not well established. Plausibility can be assessed via application of the Hill Criteria, named for a 1965 publication by Sir Austin Bradford-Hill, in which he described nine ‘viewpoints’ or ‘considerations’ by which an association described in an epidemiologic study could be assessed for causality (Hill, 1965). Hill declined to call his viewpoints ‘criteria’ lest they be considered as checklist for assessing causation. The term ‘Hill Criteria’ is used widely in the literature, however, and for convenience it is used in the present discussion. Of the nine criteria, there are seven that have utility for assessing the plausibility of an investigated specific causation relationship, as follows: . . . . .

Coherence—A causal conclusion should not contradict present substantive knowledge—it should ‘make sense’ given current knowledge. Analogy—The results of a previously described causal relationship may be translatable to the circumstances of a current investigation. Consistency—The repeated observation of the investigated relationship in different circumstances or across a number of studies lends strength to a causal inference. Specificity—The degree to which the exposure is associated with a particular outcome. Biological Plausibility—The extent to which the observed association can be explained by known scientific principles.


The attributable risk is the proportion of the total risk in the CRR that is attributable to the exposure; this is the excess risk associated with the exposure. The result of the analysis, described as either a CRR or PC, meets the legal standard of what is ‘more likely true than not’ the cause of the disease or injury in the individual when the CRR is & 2.0 (95% CI > 1.0 lower boundary), or the PC is & 50% (Federal Judicial Center, 2011). A CRR of > 1.0 and < 2.0 may be an indication that an exposure substantially contributed to the cause of the disease (if the lower boundary of the 95% CI is > 1.0), but not that it was the cause on a more probable than not basis, as the PC will be < 50%. Depending on the type of analysis underlying the CRR, the ratio may be the same as a risk ratio (relative risk) or odds ratio derived from an exposure (in an epidemiologic cohort study) or outcome (in an epidemiologic case-control study) study, respectively. For some analyses, however, the CRR is based on disparate measures of risk, and cannot be characterized as either a risk ratio or an odds ratio. An example of this circumstance was given in the hypothetical death case where the decedent had three different possible causes of death. While the risk of each of the plausible causes could be quantified with prior study, it is unlikely that there would be any studies that have compared all three causes in a similar population. Thus, the CRR is unique to the individual investigation.


5 of 10

.

Experiment—In some cases there may be evidence from randomized experiments (i.e. drug trials). . Dose response—The probability, frequency, or severity of the outcome increases with increased amount of exposure.


Subsequent authors have added the feature of Cessation/Dechallenge—Rechallenge for circumstances when the exposure is repeated over time and there is the ability to observe the associated outcome response, as might occur with an adverse reaction to a medication (Miller et al., 2000; Naranjo et al., 1981; Perrio et al., 2007). Additional considerations when assessing an association are the potential impact of confounding and bias in the data, which can obscure a true relationship. Confounding refers to a situation in which an association between an exposure and outcome is all or partly the result of a third factor that affects the outcome but is unaffected by the exposure. Bias refers to a form of error that may threaten the validity of a study by producing results that are systematically different than the true results. Two main categories of bias in epidemiologic studies are selection bias, which occurs when study subjects are selected as a result of another unmeasured variable that is associated with both the exposure and outcome of interest; and information bias, which is systematic error in the assessment of a variable. While useful when assessing a previously unexplored association, there is no combination or minimal number of these criteria that must be fulfilled in order to conclude that a plausible relationship exists between a known exposure and an observed outcome. In many cases there is no need for this first step of the assessment if a general causal relationship is well established. In large part, plausibility of a relationship is entertained once implausibility has been rejected. The two remaining Hill criteria are Temporality and Strength of Association. While both criteria have utility in assessing specific causation, temporality is the feature of an association that must be present, at least with regard to sequence (i.e. the exposure must precede the outcome), in order to consider a relationship causal. Temporal proximity can also be useful in some specific causation evaluations, as the closer the investigated exposure and the outcome are in time the less opportunity there is for an intervening cause to act. Another feature of temporality that may have a role in a specific causation evaluation is latency. An outcome may occur too soon or too long after an exposure to be considered causally related. As an example, some food borne illnesses must incubate for hours or days after ingestion, and thus an illness that begins directly following a meal, and which is later found to be caused by a food borne microorganism that requires > 12 hour incubation, was not caused by the investigated meal, even if an investigation reveals the microorganism in the ingested food. Strength of Association is the criterion that is used in general causation to assess the impact of the exposure on the population, and is often quantified in terms of relative risk. In a specific causation evaluation the strength of the association between the exposure and the outcome is quantified by the CRR, as described earlier. The two types of error that can occur with an FE assessment of specific causation are Type I, in which it is concluded that there is a causal relationship when there is not, and Type II, in which it is concluded that there is not a causal relationship when there is. A sensitivity analysis using a range of reasonable input values is recommended when the CRR is small in order to quantify potential error. A safety analysis, in which the values least favourable to the party for which the analysis is performed are used, is another means of reducing the probability of error.

6 of 10


3. Applications of forensic epidemiology 3.1 Criminal investigation Causation in criminal cases is most often straightforward because of the high degree of association between diagnostic findings associated with intentional injury and the cause of the findings. In other words, the lethality of the methods used to commit homicide (firearms, blunt trauma, sharp instruments) typically leaves little doubt about competing causes of death. In some cases, however, death and injury investigations are aided by the use of epidemiologic data or concepts; examples are abusive head trauma in infants, and whether certain diagnostic findings are reliable indicators of intentional trauma (Hymel, 2013), or whether certain injury patterns are consistent with homicide versus suicide (Freeman, 2007) or occupant position (driver or passenger) in a fatal traffic crash (Freeman et al., 2012).

Investigation of injury or disease associated with toxic exposure typically requires evidence of general causation from epidemiologic study and appropriate application of the Hill criteria pertaining to the plausibility of the association. The association may be so well established and large that it does not need to be quantified in individual cases, even if the relationship is not highly specific. An example is smoking and lung cancer; if a lifetime smoker develops lung cancer, there is no need for an FE analysis to determine the cause of the cancer. In contrast, if a 40 year-old male non-smoker develops lung cancer after 20 years of exposure to organic solvents, then an FE analysis can be helpful for assessing the probability of causation associated with the exposure. 3.3 Injury litigation Causation of injuries associated with mechanical trauma, most commonly resulting from traffic crashes and falls, is rarely an issue of dispute outside of litigation. When such injuries are evaluated in a medicolegal setting, however, speculative opinions regarding risk and probability of causation are rather common. When injury is associated with a lower speed collision, for example, it is often erroneously alleged that the low risk of injury from the collision equates with a low probability of causation (Walz and Muser, 2000). More severe collisions may involve inaccurate assertions that a particular injury pattern is a reliable indication of seat belt use or non-use (Freeman et al., 2014). Quantification and comparison of competing risks in such litigation using FE methods provides reliable evidence for fact finder determinations of cause. 3.4 Medical negligence claims The causal relationship between an alleged act of medical negligence and an adverse outcome is a pivotal element of medical malpractice litigation. An FE analysis of a medical negligence claim is often directed at the base risk of the injury as a counterfactual assessment of causation. In other words, an outcome that is plausibly caused by an alleged act of negligence is more likely to have resulted from the negligent act if the base risk of the outcome absent the negligent act is quite small. Such analyses often depend on analysis of information mined from hospital discharge databases, as well as previously published epidemiologic studies documenting the risk associated with various therapies and procedures.


3.2 Environmental toxin exposure


7 of 10

4. Case studies The following two case studies serve as illustrations of the practical application of the methods described in this report. The first case study demonstrates the application of FE methods to a set of facts from an alleged act of medical negligence, and the second case study describes the investigation of an alleged homicide. The result of both analyses was an estimate of the probability of specific causation relative to the investigated exposure, given in terms of CRR and PC. 4.1 Case study #1: Cardiomyopathy following exposure to doxorubicin


In 2007 a 26-year-old woman who was 30-weeks pregnant was diagnosed with invasive ductal carcinoma of the breast. Following successful cesarean section birth of the baby, the woman underwent a total mastectomy of the involved breast. Shortly thereafter she was started on a regimen of chemotherapy, including treatment with doxorubicin, a drug with well established cardiotoxic side effects (Singal and Iliskovic, 1998). Six months after starting on the doxorubicin the woman developed symptoms of acute decompensated heart failure, a condition that was ultimately diagnosed as resulting from a dilated cardiomyopathy. Approximately 1.5 years later she underwent a heart transplant. The allegation of medical negligence concerned the failure on the part of the physician prescribing the doxorubicin to assess the patient’s heart function prior to the initiation of the therapy, or to adequately assess her family history of cardiomyopathy, which was positive. The defence to the allegations was that competing causes of the cardiomyopathy, including an undiagnosed viral infection or idiopathic cause, as well as peripartum cardiomyopathy, could not be ruled out. The cardiology expert who provided this opinion acknowledged that the doxorubicin could have been the cause, but that she could not state which of the causes was more likely. An FE analysis was undertaken to assess the CRR of the doxorubicin exposure versus all other causes asserted by the cardiology expert. Both the doxorubicin and the viral/ idiopathic causes were deemed plausible, but the peripartum cardiomyopathy was not. A review of the diagnostic criteria for peripartum cardiomyopathy indicated that the onset of the symptoms had to be within 5 months of birth, and that there could not be any known competing causes present (Hibbard, 1999). Neither criterion was met for the case. The CRR-numerator risk assigned to the doxorubicin exposure was estimated from an analysis of the dose received by the patient, which was compared to the dose-adjusted complication rate reported in a summary of clinical trials of the drug (Swain et al., 2003). A binomial logistic regression of the reported data resulted in an estimated risk of symptomatic cardiotoxicity at the dose received by the patient of 0.85% (95% CI 0.64%, 1.13%), or approximately 1 in 118 exposed patients. In order to estimate a patient-specific CRR-denominator risk of the competing causes of the cardiomyopathy, relevant data from the Nationwide Inpatient Sample Database (NIS) of the Healthcare Utilization Project of the Agency for Healthcare Research and Quality of the U.S. Department of Health were accessed. The NIS is a publically held database containing data from approximately 8 million US hospital stays each year in 45 states, or around a 20% sample of all hospital discharges (Agency for Healthcare Research and Quality, 2014). The NIS data allow for a national estimate of the incidence, risk factors, outcomes, and other variables pertaining to all conditions seen in U.S. hospitals. Cases of viral or idiopathic cardiomyopathy diagnosed in women of the 5-year age block of the patient were identified in the NIS using appropriate diagnostic codes for data abstracted for the 2-year

8 of 10


4.2 Case study #2: Fetal death following maternal cocaine ingestion In 2006 a 15-year-old crack cocaine-using African-American female gave birth to a stillborn fetus at 37-weeks gestation. A toxicologic examination of fetal blood indicated the presence of a small and non-lethal amount of benzoylecgonine, a cocaine metabolite. Based solely on this finding the pathologist who performed the autopsy on the fetus determined that the manner of death was homicide. Based on the laws in the U.S. state where the birth occurred, the mother was charged with first-degree murder. Stillbirth occurs both with and without maternal-fetal cocaine exposure, and maternal-fetal cocaine exposure occurs both with and without stillbirth. Further, stillbirth occurs disproportionately among disadvantaged and women of colour (Stillbirth Collaborative Research Network Writing Group, 2011). Although the etiology of stillbirth in individual cases is often unclear, a number of associated factors, including poverty, single motherhood, inadequate prenatal care, maternal age, infection, obesity, diabetes, thrombophilia, fetal genetic or structural abnormalities, and umbilical cord abnormalities have been identified. An analysis of the epidemiologic literature indicated non-significant elevation of risk for stillbirth secondary to maternal-fetal cocaine exposure (Miller et al., 1995; Wolfe et al., 2005). In the FE analysis, the relationship between maternal-fetal cocaine exposure and stillbirth was considered to be plausibly causal. To further quantify the relationship a case specific analysis of hospital inpatient birth data was performed. Data from the Nationwide Inpatient Sample Database (NIS) of the Healthcare Utilization Project of the Agency for Healthcare Research and Quality of the U.S. Department of Health were accessed. Initially, a univariate analysis of the contribution of maternal cocaine presence to stillbirth risk, along with other known risk factors, was conducted. These findings were used to construct an adjusted model of the relationship between cocaine exposure and of stillbirth, using binomial logistic regression. The results of the analysis resulted in an odds ratio of 1.58 (95% CI 1.02, 2.45). This value was used as a CRR for the analysis, and was converted to a probability of causation of 37%. As a result of the FE analysis it was concluded that non-lethal maternal-fetal cocaine exposure in a case of stillbirth does not account for more than 50% of the cause of the stillbirth. The assumption by


period during which the alleged negligence occurred. Following the exclusion of cases with significant comorbidities that the patient did not have, the number of cases was matched to census data to arrive at an age and gender-specific annual rate. The result of the analysis was an estimated annual rate of viral or idiopathic cardiomyopathy in women the age of the patient of 0.018%, or 1 in 5640 women. Adjusted to the period of 6 months from the date of the initiation of the therapy to the date of the first symptoms of heart failure, the risk applicable to the circumstances of the case would be half of the annual rate: approximately 0.009%, or 1 in 11 280. A CRR using the drug-related cardiotoxicity risk of 1 in 118 versus the competing 6-month risk of a randomly occurring cardiomyopathy of 11 280 was equal to a ratio of 95.6 to 1 in favor of the doxorubicin as the cause of the patient’s cardiomyopathy. The estimate was considered statistically significant, as the lower bound of the confidence interval did not include 1.0. Using the formula described earlier, this CRR is equal to a probability of causation of approximately 99%. As a result of the analysis, it was concluded that the probability of cause of the patient’s cardiomyopathy attributable to the doxorubicin exposure versus the only other plausible cause that was considered in the case during the time of exposure was approximately 99%.


9 of 10

the pathologist that the presence of fetal cocaine was highly specific for the stillbirth, and thus that the manner of death was homicide rather than due to natural causes, was rejected as erroneous. While the cocaine exposure could have caused the stillbirth, it could not be concluded that the exposure was the most probable cause of the stillbirth, much less that a homicide had been committed beyond a reasonable doubt, which is the relevant standard of proof for a criminal conviction. 5. Conclusion

Disclosures Both authors provide medicolegal consultation services. REFERENCES Agency for Healthcare Research and Quality. 2014. http://hcup-us.ahrq.gov/nisoverview.jsp (accessed 10 January 2015). COLVILLE-EBELING B, FREEMAN M, BANNER J, LYNNERUP N. Autopsy practice in forensic pathology - evidence-based or experience-based? a review of autopsies performed on victims of traumatic asphyxia in a mass disaster. J Forensic Leg Med 2014; 22: 33–6. Federal Judicial Center. Reference manual on scientific evidence. 3rd ed. Washington, DC: National Academy Press; 2011. FREEMAN M. Bayesian analysis of predictive characteristics in suicidal versus homicidal hanging deaths: a case study in forensic epidemiology. In: 59th annual meeting of the American Academy of Forensic Sciences; 2007 February 19–24; San Antonio, Texas: American Academy of Forensic Sciences; 2007. p. 304. FREEMAN MD, DOBBERTIN K, KOHLES SS, UHRENHOLT L, ERIKSSON A. Serious head and neck injury as a predictor of occupant position in fatal rollover crashes. Forensic Sci Int 2012; 222(1–3): 228–33. FREEMAN MD, ERIKSSON A, LEITH W. Injury pattern as an indication of seat belt failure in ejected vehicle occupants. J Forensic Sci 2014; 59(5): 1271–4. FREEMAN MD, ROSSIGNOL AM, HAND ML. Forensic epidemiology: a systematic approach to probabilistic determinations in disputed matters. J Forensic Leg Med 2008; 15(5): 281–90. HIBBARD JU, LINDHEIMER M, LANG RM. A modified definition for peripartum cardiomyopathy and prognosis based on echocardiography. Obstet Gynecol 1999; 94(2): 311–6. HILL AB. The Environment and disease: association or causation? Proc R Soc Med 1965; 58: 295–300 HOLLAND P. Statistics and causal inference. Am J Stat Ass 1986; 81(396): 945–60. HYMEL KP, WILLSON DF, BOOS SC, PULLIN DA, HOMA K, LORENZ DJ, et al. Derivation of a clinical prediction rule for pediatric abusive head trauma. Pediatr Crit Care Med 2013; 14(2): 210–20. KOEHLER SA, FREEMAN MD. Forensic epidemiology: a method for investigating and quantifying specific causation. Forensic Sci Med Pathol 2014;10(2): 217–22. MILLER JM, JR., BOUDREAUX MC, REGAN FA. A case-control study of cocaine use in pregnancy. Am J Obstet Gynecol 1995; 172(1 Pt 1): 180–5. MILLER FW, HESS EV, CLAUW DJ, HERTZMAN PA, PINCUS T, SILVER RM, et al. Approaches for identifying and defining environmentally associated rheumatic disorders. Arthritis Rheum 2000; 43(2): 243-9.


The lack of a broad understanding and adoption of a systematic and scientifically valid approach to the evaluation of specific causation in medicolegal settings is a potential source of erroneous opinion. Forensic epidemiology is a branch of forensic medicine that provides a scientifically valid basis for fact finder determinations of cause and effect when such relationships are not readily apparent or in dispute.

10 of 10



NARANJO CA, BUSTO U, SELLERS EM, SANDOR P, RUIZ I, ROBERTS EA, et al. A method for estimating the probability of adverse drug reactions. Clin Pharmacol Ther 1981; 30(2): 239–45. National Research Council. Strengthening forensic science in the United States : a path forward. Washington, DC: National Academies Press; 2009. Report No.: 9780309131353. O’NEAL JT. Causation determination in workers’ compensation and toxic tort cases. 2001. http://www.health. state.mn.us/divs/eh/hazardous/topics/eegrandrounds/causation.pdf (accessed 12 January 2015). PERRIO M, VOSS S, SHAKIR SA. Application of the bradford hill criteria to assess the causality of cisapride-induced arrhythmia: a model for assessing causal association in pharmacovigilance. Drug Saf 2007; 30(4): 333–46. POLLANEN MS. Forensic pathology and the miscarriage of justice. Forensic Sci Med Pathol 2012; 8(3): 285–9. ROTHMAN KJ, GREENLAND S. Causation and causal inference in epidemiology. Am J Public Health 2005; 95 Suppl 1: S144–50. SINGAL PK, ILISKOVIC N. Doxorubicin-induced cardiomyopathy. N Engl J Med 1998; 339(13): 900–5. Stillbirth Collaborative Research Network Writing Group. Causes of death among stillbirths. JAMA 2011; 306(22): 2459–68. SWAIN SM, WHALEY FS, EWER MS. Congestive heart failure in patients treated with doxorubicin: a retrospective analysis of three trials. Cancer 2003; 97(11): 2869–79. WALZ FH, MUSER MH. Biomechanical assessment of soft tissue cervical spine disorders and expert opinion in low speed collisions. Accid Anal Prev 2000; 32(2): 161–5. WOLFE EL, DAVIS T, GUYDISH J, DELUCCHI KL. Mortality risk associated with perinatal drug and alcohol use in California. J Perinatol 2005; 25(2): 93–100.