On November 20, 2021, petroleum jet propellant-5 (JP-5) fuel was released from a fire-suppression line, eventually leaking into the Joint Base Pearl Harbor Hickam (JBPHH) Water Distribution System on Oahu, Hawaii and contaminating the water supply for an estimated 93,000 people. This system predominantly serves military residences and buildings (i.e., offices, schools), although some civilians were also exposed. Approximately one week later, the Hawaii State Department of Health (HDOH) began receiving complaints of a fuel-like odor from the tap water among consumers of the JBPHH water system, and on November 29, 2021, issued a drinking water advisory for consumers to avoid using the water for drinking, cooking, bathing, dishwashing, laundry or oral hygiene (i.e., brushing teeth). Literature is limited on potential long-term human health effects from exposure to JP-5, relying on studies of other jet fuels.^1–4 HDOH therefore requested the U.S. Centers for Disease Control and Prevention (CDC)/Agency for Toxic Substances and Disease Registry (ATSDR) conduct Assessment of Chemical Exposures (ACE) investigations to assess health effects associated with the contamination event.^5,6 On March 18, 2022, HDOH amended the drinking water advisory to safe for all uses. Six months after that amendment, a follow-up ACE investigation found that 75% of participants were still reporting ≥1 symptoms in the past 30 days.⁶ That investigation recommended a detailed medical review to evaluate ongoing health effects of potential JP-5 exposure. This might help guide decisions on health services needed in this community and determine the necessity for follow-up studies. Thus, the U.S. Department of Defense (DoD), Defense Health Agency (DHA) requested CDC/ATSDR assistance in reviewing medical charts within the military health system (MHS) to evaluate continued health concerns in the community served by the contaminated water.

The CDC/ATSDR team reviewed electronic health records (EHRs) at Tripler Army Medical Center from February 24 to March 17, 2023. Each team member completed training regarding electronic health system use and best practices for patient confidentiality. DHA subsequently provided restricted on-site access to EHRs. Overall record abstraction was time-constrained due to completing necessary tasks upon arrival to Oahu, including gaining MHS familiarity and designing a REDCap database to store abstracted data.

The water contamination event was defined as occurring on the fuel release date of November 20, 2021, and the exposure window continued through March 18, 2022, when HDOH fully amended the drinking water advisory. A roster of persons served by the contaminated water source during the exposure window was obtained from the Defense Occupational and Environmental Health Readiness System (DOEHRS)⁷. It was used to identify persons who sought medical care through the military health networks during November 20, 2021–February 24, 2023. Persons were convenience sampled from the roster based on non-mutually exclusive subgroups of International Classification of Diseases, Tenth Revision diagnostic codes (Supplemental Table 1). This included codes for toxic effects and symptoms of clinical interest. Additionally, we selected from all eligible pregnancy-related codes during November 20, 2021–June 18, 2022, and 10% random samples of pediatric (ages 0 to <18 years old) and adult persons with codes from multiple visits.

Deidentified data, including demographics, medical history, laboratory and imaging results, medications, vital signs, referrals and visit summaries, and pediatrics and pregnancy information, were abstracted into the REDCap database. The final sample included records with limited missing demographic information. Medical histories included conditions defined as recorded clinical diagnoses, and symptoms defined as manifestations of diseases recorded from patient complaints. Our case definition categorized conditions and symptoms as the following: preexisting, if documented before November 20, 2021; worsening preexisting, if documented as present before November 20, 2021 with evidence of worsening in >2 visits after November 20, 2021; persistent new, if documented on or after November 20, 2021 and lasting ≥3 months; or acute, if first documented on or after November 20, 2021 and lasting for <3 months. Persons with conditions or symptoms classified as worsening preexisting or persistent new were highlighted in this brief report to better understand potential chronic comorbidities. Laboratory and vitals data conducted during clinical visits were collected beginning November 20, 2020, to establish a baseline and help determine worsening preexisting conditions. Pregnancy records included pregnancy outcomes for singleton or multiple births (i.e., livebirth [full-term birth ≥37 weeks or preterm birth <37 weeks], miscarriage, stillbirth, and loss to follow-up). To preserve patient confidentiality and reduce misinterpretation of unstable data, we suppressed any count data with less than five persons reported except when data is reported as missing.

This activity was reviewed by CDC, deemed not research, and was conducted consistent with applicable federal law and CDC policy (45 C.F.R. part 46.102 [2], 21 C.F.R. part 56; 42 U.S.C. Sect. 241[d]; 5 U.S.C.0 Sect.552a; 44 U.S.C. Sect. 3501 et seq.); consenting was not required under DoD Manual 6025.18.

With the limited field time, full medical chart abstraction was completed for 653 of 28,601 persons on the roster. We included 405 (62%) persons with codes for toxic effects, and 170 (26%) persons with codes of clinical interest. We abstracted records for 239 (37%) persons ages 0–17 years and 45 (7%) pregnancies. Of the 653 persons, most were adults (63%; 411/653), female (55%; 359/653), and identified as White (49%; 319/653) or non-Hispanic (64%; 420/653) [Table 1]. Most were military dependents (i.e., not active duty military but able to access military healthcare services due to affiliation, such as spouse or child of an active duty military member) (64%; 416/653); the most-affiliated service branches were Army (35%; 231/653), Navy (31%; 200/653), and Air Force (27%; 175/653). Most military service members were enlisted (82%; 172/211), as opposed to commissioned officers who are of higher rank and require a higher education for their commanding roles. Approximately two-thirds (69%; 449/653) had documentation (e.g., response clinic intake form) of exposure to the water contamination event within EHRs.

Approximately half (55%; 357/653) were identified with worsening preexisting or persistent new conditions or symptoms. The most frequently documented worsening preexisting conditions included migraine (8%; 50/653), chronic pain (4%; 26/653), and gastroesophageal reflux (3%; 16/653); most frequently documented persistent new conditions included migraine (2%; 14/653), adjustment disorder (2%; 13/653), and eczema or atopic dermatitis (2%; 12/653) (Table 2). The most frequently documented worsening preexisting symptoms included headache (8%; 49/653), anxiety (6%; 42/653), and depression or low mood (4%; 23/653); most frequently documented persistent new symptoms included rash (7%; 46/653), headache (5%; 34/653), and fatigue or malaise (4%; 29/653). Adults more commonly had documented neurological (e.g., migraine or headache) or neuropsychological (e.g., anxiety or adjustment disorder) conditions and symptoms, whereas children (ages 0 to <18 years old) more commonly had dermatological (e.g., eczema or rash) or gastrointestinal (e.g., abdominal pain) conditions and symptoms.

Of 45 pregnant women, 38% (17/45) had worsening preexisting and 38% (17/45) had persistent new conditions or symptoms. In total, 80% (36/45) experienced a singleton full-term birth, and 13% (6/45) experienced a miscarriage. Outcomes for those experiencing a singleton preterm birth and lost to follow-up after 20 weeks gestation had counts less than five and were thus not reported.

Exploratory results in this pilot investigation demonstrate certain chronic conditions worsened and new conditions persisted after the November 2021 water contamination event. Most persons in our sample of records did not develop a condition during the study period. We observed certain impaired neurological functions, such as migraine-related sensory impairments (i.e., vision, movement/vertigo/dizziness) and fatigue or malaise, and neuropsychological functions, such as anxiety, attention deficit/hyperactivity disorder, depression, mental health and psychiatric disorders, and gastrointestinal distress after potential exposure to JP-5.

This is consistent with previous, but limited, literature examining exposure to hydrocarbons, such as JP-8.^3,8,9 Although we were unable to systematically confirm any direct routes of exposure within EHRs, we hypothesized multifactorial exposure routes, including ingestion and dermal absorption, given the range of conditions and symptoms presented. Aside from direct exposure to contaminated water, other aspects of this event, such as housing relocation, distrust of systems, and stress, might inform these observed health effects.¹⁰ Of note, this water contamination exposure event occurred during the COVID-19 pandemic, which might have influenced some conditions or symptoms experienced.

This investigation used convenience sampling, which may introduce selection bias, and was unable to quantify exposure, establish a dose-response, or determine causal associations between exposure to the contamination event and health outcomes experienced. The DOEHRS roster only includes DoD-affiliated individuals, and we were only able to assess 2% of the roster over the course of three weeks. Regardless, the exposure occurred through the Navy water system which predominantly serves DoD-affiliated households. We strategically sampled from the potentially exposed population based on toxic effects and comprehensively evaluated EHRs to understand overall prevalence of conditions and symptoms. Information bias might be present and attributable to data availability or collection methods, variations in patient reporting behaviors or provider charting, or abstraction procedures. EHRs and patient tracking were limited by availability within MHS. However, we largely did not have access to records for purchased care, those who were referred or sought care outside of this network. Future analyses might use purchased-care records to incorporate data from specialized referrals outside of this network, as affected persons might seek external care because of distrust or lack of confidence in military providers.

Approximately one-third of reviewed charts lacked documentation of exposure. Among persons with documented exposure, the type and extent of documentation varied. During disaster or emergency responses, data should be systematically collected using standardized resources and methods. Misclassification can occur during abstraction when synthesizing and performing a secondary analysis on EHRs. To minimize subjective clinical impressions, the team was co-located during abstraction to allow for routine consultations spanning our breadth of clinical and epidemiological specialties.

Approximately half of abstracted records had documented worsening preexisting or persistent new conditions or symptoms after the water contamination event. This investigation highlights the need to continue addressing the burden of morbidities among this population and extensively monitor persons affected for long-term health outcomes. Future efforts might focus on effects of exposure among sensitive populations, particularly children and pregnant women, and comparison to a control population to better understand the nature of any increased risk among exposed persons and to assess the impact on the overall population. Access to medical care, including mental health resources and specialty providers, and monitoring for long-term health outcomes among potentially exposed communities will be important.