Estimating Acute Viral Hepatitis Infections From Nationally Reported Cases

Abstract

Objectives. Because only a fraction of patients with acute viral hepatitis A, B, and C are reported through national surveillance to the Centers for Disease Control and Prevention, we estimated the true numbers.

Methods. We applied a simple probabilistic model to estimate the fraction of patients with acute hepatitis A, hepatitis B, and hepatitis C who would have been symptomatic, would have sought health care tests, and would have been reported to health officials in 2011.

Results. For hepatitis A, the frequencies of symptoms (85%), care seeking (88%), and reporting (69%) yielded an estimate of 2730 infections (2.0 infections per reported case). For hepatitis B, the frequencies of symptoms (39%), care seeking (88%), and reporting (45%) indicated 18 730 infections (6.5 infections per reported case). For hepatitis C, the frequency of symptoms among injection drug users (13%) and those infected otherwise (48%), proportion seeking care (88%), and percentage reported (53%) indicated 17 100 infections (12.3 infections per reported case).

Conclusions. These adjustment factors will allow state and local health authorities to estimate acute hepatitis infections locally and plan prevention activities accordingly.

Infection with hepatitis A, B, and C (HAV, HBV, and HCV, respectively) remains a substantial health problem in the United States.^1–3 Chronic HBV⁴ and HCV⁵ infections currently affect more than 4 million US residents and now account for more deaths than does HIV/AIDS.⁶ New HAV and HBV infections have been prevented by the adoption of universal infant vaccination in the United States, but acute infections continue to occur and cause substantial morbidity and mortality.⁷ Monitoring case patients with acute HAV, HBV, and HCV is important for several reasons. Identifying individuals with acute infections serves to describe modes of transmission and to detect and control outbreaks. Furthermore, prevention interventions of various types—for example, vaccinating susceptible persons, getting injection drug users into treatment programs, treating persons who are chronically infected to prevent secondary transmission, and preventing complications—require ongoing surveillance and analysis of individuals with acute infections of these 3 viruses.

Only a fraction of individuals with acute infections of these 3 viruses are reported eventually to the Centers for Disease Control and Prevention (CDC) in the United States. Barriers to ascertaining and reporting hepatitis infections are many, often reflecting the ability of and resources allotted to the local and state health jurisdictions monitoring them. Natural barriers include the following: most individuals with acute infections of any of the 3 viruses are asymptomatic, only some of those with symptoms seek medical care and testing, and even of those diagnosed, some fraction is not reported or enumerated. Complete reporting of, at a minimum, symptomatic case patients is essential because only by identifying them can interventions be implemented to limit disease in the community. Outbreaks are most often detected from the identification of symptomatic case patients.

Currently, the CDC estimates incident HAV, HBV, and HCV infections using reports of case patients to develop adjustments through 3 simple probabilistic multiplier models; however, the methods used to develop the estimating factors (multipliers) are outdated and have never been well described or publicly available. Thus, our goal was to update estimates of the number of individuals with acute infections using 2011 reports of case patients.

METHODS

We adapted a model used previously to estimate the number of those infected with pandemic influenza in 2009. That model is relevant because hepatitis and influenza virus infections have similar surveillance challenges, including the following: many infected individuals are asymptomatic, not all ill persons seek care, specimens sometimes yield false results, and not all true case patients are reported to health departments.⁸ Our model used a simple multiplier approach requiring 3 parameters: (1) the proportion of infected persons developing symptoms, (2) the proportion of those who sought care, and (3) the proportion of those diagnosed who were reported to hepatitis surveillance (Figure 1). We obtained these parameter proportions from a literature search. We included the proportion from each of the studies in a random effects model, resulting in a pooled proportion.⁹ Then, we multiplied the inverse of the pooled proportions by the number of case patients reported to the National Notifiable Diseases Surveillance System (NNDSS)⁷ to derive the final estimates.

FIGURE 1—

Steps used to estimate new infections of viral hepatitis among adults: United States, 2011.

Study Selection and Data Extraction

We searched PubMed for studies of acute HAV, HBV, or HCV using combinations of the following medical subject headings terms: “hepatitis virus,” “disease outbreaks,” “human hepatitis A virus,” “hepatitis B,” “acute hepatitis C,” and “United States.” We then abstracted the number of individuals with acute infections identified and the proportion exhibiting signs or symptoms.

In addition, we searched among outbreak investigations for which adequate records allowed the calculation of symptomatic case patients among acutely infected persons. We obtained these from records of the health care transmission of hepatitis since 1989 and published online.¹⁰

Inclusion Criteria

We included studies of US adult populations that were published in English from 1980 through October 2012. We included only studies that used serology to define new infections. Some studies used signs and symptoms as their case definition, and we did not include those reports.

None of the studies described the frequency with which persons with viral hepatitis symptoms sought care. Instead, we used data from the 2010 National Health Interview Survey.¹¹ Briefly, the National Health Interview Survey uses a probability sampling of US households. The questions we analyzed were asked of adults aged 18 years or older: “Is there a place that you usually go to when you are sick or need advice about your health?” and “Have you ever had hepatitis?” We calculated the prevalence of responding yes to both questions: a person with hepatitis infection reporting a routine source of care.

Source of Reported Case Patients

In the United States, symptomatic case patients with acute hepatitis infections are reported to the NNDSS. State and territorial health departments send reports voluntarily every week. Case patient definitions for acute HAV, HBV, and HCV are established in collaboration with the Council of State and Territorial Epidemiologists and require a combination of clinical and laboratory criteria. Definitions are available online.¹²

We used data reported to the NNDSS in 2011 to identify the number of case patients reported for the multiplier approach.⁷ For HCV only, we estimated the number of reported individuals who may have been infected by injection drug use¹³ because 78% of case patients with acute HCV reported to NNDSS have missing or no information on mode of transmission.⁷

RESULTS

Our literature search identified 226 studies published since 1980. Of these, 39 met our inclusion criteria (Table 1). Of the outbreak investigations, most were captured in the literature review; only 1 that had measured the frequency with which infected persons presented signs or symptoms had not been published in a peer-reviewed journal.⁵¹ There were 6 studies of HAV, 15 of HBV, and 13 of HCV that described the proportion of infections with signs or symptoms of hepatitis. Even fewer studies measured underreporting of symptomatic infections to surveillance; we found 2 for HAV and 3 each for HBV and HCV.

TABLE 1—

Model Parameters and Sources of Data Used to Adjust Number of Case Participants Reported to Surveillance for Estimating Incident Viral Hepatitis Infections: United States, 2011

Parameter/Setting	Year	Location and Reference	No./Total (%)	Pooled Estimate, %
Hepatitis A
Frequency of symptoms				84.8
Restaurant staff	2003	Pennsylvania14	13/13 (100)
Health department	2005	Connecticut15	108/127 (85)
Health department	2005	Alaska15	27/37 (73)
Sentinel Counties Study of Viral Hepatitis	2005	US counties15	53/140 (38)
Military field training	1982	US military base, Germany16	28/29 (97)
State legislators	1981	Tennessee17	7/8 (88)
Country club members	1978	Minnesota18	97/102 (95)
Residents of a trailer park	1982	Bartow County, GA19	31/35 (89)
Military prison	1982	Kansas16	35/46 (76)
Sought care: noninstitutionalized, household population	2012	United States20	(88)	88.0
Underreporting to health department				68.7
Health department	1987	Pierce County, WA21	14/15 (93)
Managed care organization	2008	Massachusetts22	1/4 (25)
Hepatitis B
Frequency of symptoms				39.5
Weight loss clinic	1986	California23	27/60 (45)
Pain clinic outbreak	2004	Oklahoma24	13/31 (42)
Injection drug users	2003	Natrona County, WY25	27/45 (60)
State correctional facility	2000	Undisclosed26	2/6 (33)
Inpatient surgeries	1992	Undisclosed27	6/19 (32)
2 dialysis centers	1996	Allegheny County, PA28	0/6 (0)
Private physician office	2005	New York, NY29	4/38 (11)
Assisted living facility	2010	Illinois30	7/8 (88)
Orthopedic surgeon office	2010	Virginia31	1/2 (50)
Injection drug users	2002	Seattle, WA32	14/63 (22)
Injection drug users	2004	Montana33	11/21 (52)
Nursing home	2004	Mississippi34	2/15 (13)
Nursing home	2003	North Carolina34	1/11 (9)
Hematology oncology clinic	2009	New Jersey35	16/19 (84)
Outpatient gastroenterology facility	2005	New York, NY36	2/6 (33)
Nursing home	2012	North Carolina37	3/6 (50)
Nursing home	2012	North Carolina38	0/6 (0)
Psychiatric long-term care	2012	Cook County Illinois38	5/8 (63)
Ambulatory psychiatric residents	2008	California39	6/9 (67)
Sought care: noninstitutionalized, household population	2012	United States20	(88)	88.0
Underreporting				44.5
Active surveillance of population	1987	Pierce County, WA21	19/27 (70)
Managed care organization	2008	Massachusetts40	4/8 (50)
Injection drug users	2002	Seattle32	2/14 (14)
Hepatitis C
Frequency of symptoms, injection drug users				12.8
Injection drug users	2011	Massachusetts41	4/28 (14)
Injection drug users	2002	Seattle32	4/53 (8)
Injection drug users	2004–2007	New York State42	4/20 (20)
Frequency of symptoms, non–injection drug users				47.6
Pain clinic outbreak	2004	Oklahoma24	40/71 (56)
Pain clinic outbreak	1999–2002	Oklahoma43	27/47 (57)
Hematology and oncology clinics	2001	Nebraska44	16/99 (16)
Cardiology and radiology clinics	2006	Maryland45	15/16 (94)
Medical facilities, blood donors	2007	Multiple US sites46	24/67 (36)
Dialysis clinic	2006	Virginia47	3/7 (43)
Research agency patients	2011	National Institutes of Health, MD48	20/25 (80)
HIV-infected men	Unknown	New York, NY49	3/11 (28)
Outpatient gastroenterology facility	2005	New York, NY36	3/6 (50)
Outpatient dialysis facility	2012	Georgia19	0/4 (0)
Sought care: noninstitutionalized, household population	2012	United States20	(88)	88.0
Underreporting				53.0
Active surveillance of a population	1987	Pierce County, WA21	16/38 (42)
6 health departments	2008	CO, CT, MN, OR, New York City, New York State50	102/120 (85)
Injection drug users	2002	Seattle, WA32	0/4 (0)

In 2010, there were an estimated 262 000 persons with viral hepatitis who also reported a usual (regular) source of medical care. The prevalence of these events together was 3.2% (95% confidence interval [CI] = 2.9%, 3.5%) among all respondents. Among persons reporting a hepatitis infection, 88.0% (95% CI = 84.5%, 91.6%) said they had a usual source of care. The inverse of this proportion (1.14) is presented (sought care) in Table 2.

TABLE 2—

Multipliers Used to Estimate the Number of New Infections With Hepatitis A, B, and C: United States, 2011

Hepatitis Type	Multiplier			Actual Case Patients Reported 2011	Estimated Number of Infectionsa	Final Multipliers (Infection–Case Ratio)
Symptomatic	Sought Care	Reported to Surveillance
A	1.18	1.14	1.46	1398	2730	1.95
B	2.54	1.14	2.25	2890	18 730	6.48
C						12.30
Injection drug use	7.81	1.14	1.89	937	15 770
Other risk factor	2.10	1.14	1.89	293	1330

^aRounded to the nearest 10.

There were 1398 case patients with HAV, 2890 with acute HBV, and 1230 with acute HCV reported to NNDSS during 2011 (Table 2). Of the case patients with acute HCV reported, we assigned 937 (76%) as having injected drug use as a risk factor; we assigned the remaining 293 to a source of infection other than injection drug use (e.g., receipt of unscreened blood, infection in health care settings).

The inverse values of the pooled estimates and resulting multipliers from the studies are presented in Table 2. For 2011 we estimated 2730 HAV infections, or 1.95 new infections for each case patient reported to the CDC. For HBV, we estimated 18 730 infections, or 6.48 infections per reported case patient. For HCV, we estimated 17 100 infections, or 12.30 infections per case patient with an acute infection.

DISCUSSION

We found that there were an estimated 38 560 new patients with acute HAV, HBV, or HCV infections during 2011. That is, for every reported individual with HAV, there were an estimated 2.0 new infections; for every reported individual with acute HBV, there were actually about 6.5 new infections; and for every reported individual with acute HCV, about 12.3 new infections. These estimates of numbers of acute infections are essential for health departments to evaluate the success of prevention activities, including vaccination for HAV and HBV, infection control practices in health care settings, and education and needle exchange and drug treatment programs for injecting drug users.

There may be other methods to estimate the number of all acute infections. For example, estimates of acute HCV infections could be derived from the National Health and Nutrition Examination Survey by testing all specimens, about 5000 per year, for HCV RNA⁵²; however, the extensive resources needed to conduct such testing are not available, and the small numbers generated would still lead to very wide statistical CIs on the estimate. Furthermore, these numbers probably would not be useful for state and local health departments.

These updated adjustment factors are somewhat lower than are those the CDC used in the past.⁷ However, previous methods used to estimate infections, as published in annual national surveillance reports, had not been updated or recalculated for many years and depended heavily on knowledge of the epidemiology on the basis of small cohorts of transfusion recipients in the 1990s.⁵³ We included recent literature and outbreaks, so our multipliers are more likely to indicate the current number of new infections.

Case reports sent electronically to CDC are useful for examining gross trends in acute HAV, HBV, and HCV infections; all of these have been declining markedly in the past 10 years.¹⁰ Specifically, HAV declined 87%, HBV declined 63%, and HCV declined 25% since 2001.⁷ HAV and HBV are thought to result from increasing vaccination coverage of the population, especially the youngest age groups. However, the steep decline in the number of case patients with acute HCV over the past 2 decades is more difficult to explain but may be related to changes in behavior among injecting drug users, screening the blood supply since 1992, and changes in standard precautions—many of these in reaction to the HIV/AIDS epidemic in the United States.⁵⁴

In 2010, the Institute of Medicine recommended removing the criterion for symptoms from the case definitions for acute HBV and HCV.⁵⁵ However, laboratory indicators for acute infection are available only for HAV and HBV (immunoglobulin M antibodies), not for HCV, and many individuals with acute infections of all 3 viral hepatitides are determined from overt symptoms, such as jaundice. To be tested, an infected person must seek care; asymptomatic persons might not seek care. Thus, removing symptoms from the case definitions would not allow back-calculation as performed in our analysis.

These estimates are limited mainly in that we had to impute the estimates for 2 criteria—proportion of symptomatic persons seeking care and then reported to local and state health departments—from indirect and sparse data. Similarly, the distribution of injecting drug users among case patients with acute HCV is from a small study (n = 21). We could not determine whether the frequency of responding to the 2 questions in the National Health Interview Survey represented an over- or underestimate of the frequency with which persons with hepatitis actually sought care because no published studies measured that event. There have been few studies directed to assessing the underreporting of persons diagnosed with acute viral hepatitis. We did not measure variability or uncertainty in the model parameters, but we expect to quantify these as a next step. Thus, these estimates should be viewed as the best that can be generated at this time to support surveillance, planning, and prevention activities until better data sets or more sophisticated models can be developed.

Many acute infections of HAV, HBV, and HCV continue to occur. The estimates and methods we have described can be used at the national, state, and local levels to estimate burden and control and prevent new hepatitis infections. Continued monitoring of individuals with symptomatic infections and periodic estimation of all infections are essential to understanding the magnitude of the problem and evaluating the impact of prevention.