Postmortem submergence interval (PMSI) and human decomposition in anthropogenically constructed aqueous environments (pools, bathtubs, hot tubs, and spas)

Abstract

Postmortem decomposition changes of bodies in aquatic environments may offer valuable insights into the postmortem submergence interval (PMSI) for medicolegal death investigators. However, the effects of immersion on the onset of such changes are poorly understood. The aims of this study were to augment what is known about the statistical relationship between PMSI and observable decomposition changes across three body regions (head, torso, and limbs) and to quantify the influence of extrinsic (e.g., clothing, immersion, location, and body position) and intrinsic (e.g., height, weight, BMI) moderating covariates for bodies in aqueous anthropogenically constructed environments. This study retrospectively examined 79 deceased adults recovered from pools, bathtubs, hot tubs, and spas between January 2010 and December 2019 in Victoria, Australia. Mean estimated PMSI was 18.33 h (SE = 3.83), and bivariate t‐tests and ANOVA showed that PMSI was significantly related to gaseous distension of the face and torso, discoloration of the skin, marbling, skin sloughing, and discoloration of the torso and limbs. Univariate GLM (two‐way factorial ANOVA) results showed that each moderating factor had a unique profile of influence on the various relationships between PMSI and the decomposition change being evaluated. Multiple linear regression modeling demonstrated the importance of including only specific moderators in estimations of PMSI. One model predicted nearly two‐thirds of the variance in PMSI (adjusted R ² = 0.656, SE of estimates = 18.150), but more cases involving lengthier submergence intervals are needed to understand with greater confidence the onset of advanced decomposition changes in such aqueous settings.

Highlights.

• PMSI and the onset of certain taphonomic changes are statistically correlated.

• Water immersion may delay the onset of decomposition.

• Taphonomic change is moderated by situational, demographic, and anatomical size factors.

• The stepwise multiple linear regression model of PMSI had an adjusted R ² = 0.618.

1. INTRODUCTION

Human decomposition has been widely studied in terrestrial contexts worldwide [1, 2, 3, 4]. A concerted research effort has also been made in the past 10–15 years to expand this knowledge to include aquatic decomposition, particularly in marine and naturally occurring water environments [1, 5, 6, 7, 8, 9]. However, these studies were conducted in ecological and environmental niches that limit their generalizability. To bolster what is known about aquatic decomposition, retrospective examinations in anthropogenically constructed environments may offer a more controlled setting to examine the effects of postmortem submergence interval (PMSI) on human decomposition. The PMSI refers to the amount of time an individual has been submerged after death; this differs from the postmortem interval (PMI), which refers to the amount of time elapsed between death and the time when the body was discovered [4]. PMSI is of forensic importance because it may help medicolegal investigators discern the approximate time of water entry and offer insights as to the postmortem journey of remains, including decomposition pathways prior to recovery. Moreover, despite major advancements in the application of insects and algae for PMI estimations in aquatic environments [10, 11, 12, 13, 14], naturally occurring water systems that allow for algae growth differ from those of indoor bathtub and pool settings, the environmental context of this study. Many existing models of estimated elapsed time since death and stages of decomposition also cannot be generalized to aqueous environments because they were constructed with terrestrial data [15] and immersion in water has been found to delay the typically expected taphonomic changes [16], particularly in waters with cooler temperatures and anaerobic environments. For these reasons, it is imperative that researchers continue to advance what is known about aquatic taphonomy.

1.1. Common aquatic decomposition patterns and moderators

Much of the existing literature on human body decomposition derives from a combination of experimental work using animal proxies [6, 17, 18, 19, 20, 21] and retrospective case studies [22, 23, 24]. Some have sought to develop aquatic decomposition scoring methods or matrices and others have been empirical studies to investigate the complex relationship between decomposition, pathology, and the environment [1, 25, 26, 27, 28, 29]. A variety of cutaneous changes, including water‐related skin maceration of the hands and feet (i.e., Waushaut or washerwoman's hands), artefactual cutis anserina from the rigor of arrector pili muscles within the skin, vascular marbling, discoloration, bloating, and sloughing of the skin have been observed in association with prolonged submergence [16]. These changes may also be seen in terrestrial decomposition [4, 15], but the differences in their onset in different environments requires further research. Retrospective aquatic studies have pointed to the need for further examination of how circumstantial data may influence taphonomic findings [22, 30, 31]. In one study, organ weight and decomposition was no longer statistically significant when corrected for BMI, highlighting the need to consider the deceased's height and weight [22]; authors also hypothesized that water slowed the decomposition process [22].

There are extrinsic factors that may influence the rate and onset of taphonomic change, including situational factors and the environment where the body decomposes [8, 25, 32]. The position of the deceased's body is an example of an extrinsic moderating factor that may influence decomposition. For instance, the body position and whether it is fully or partially immersed may influence its accessibility to scavengers, opportunistic colonizers, and saprophytes, including, but not limited to, insects and arthropods, fish, crustaceans, and barnacles [5, 8, 33, 34, 35, 36]. In anthropogenically constructed pools or indoor environments like bathtubs, the only realistic scavengers and colonizers in proximal distance to the body may be saprophytes (plant, fungus, or microorganisms that live on dead or decaying matter) or insects [10, 11, 12, 13, 37, 38, 39, 40, 41, 42, 43], which should be considered for their value as forensic entomology evidence and have been used successfully in casework [19, 44]. Wet clothing from 4 to 10 min of immersion functioned like dry clothing to prevent heat loss in cadavers placed in both indoor and outdoor settings [45], and in riverine recovered remains, clothing produced air pockets that influenced the body's position and float potential during the earlier stages of PMSI [46]. As for intrinsic factors, age, sex, and circumstances of the death have also been tested as moderators. One example includes a study on macromorphological findings in water‐related deaths across PMSI [47], which examined both internal and external signs of drowning (e.g., foam from bronchi, trachea, larynx, to mouth/nostrils) and general signs of putrefaction (as measured by bloating). Other studies have broadly discussed how clothing, immersion (fully or partially), and size of the body may influence the onset of decomposition changes [1, 5, 16, 33, 34], but few were able to quantify the effect of such moderating extrinsic factors.

1.2. Standards for the study of taphonomic observations

To organize and contextualize what aquatic decomposition changes occur, standards have been developed by major forensic organizations, and aquatic decomposition scoring (ADS) methods have been devised to better evaluate taphonomy [25, 26, 29]. The American National Standards Institute (ANSI)/Academy Standards Board (ASB) Standard 149: Standard for Taphonomic Observations in Support of the Postmortem Interval (2022) [48] calls for an interdisciplinary approach when possible. It recommends that practitioners document observations of taphonomic conditions and the taphonomic changes that result from abiotic, biotic, chemical agents, thermal events, anthropogenic effects, and decomposition processes, with a clear distinction between taphonomic events and alterations from trauma or injury (Standards 147 Subsection 4.3) [48]. It also calls for the reporting of observed factors that may influence decomposition, which can include extrinsic ecological factors, physical characteristics of the deceased, or modifiers of remains [48].

1.3. Aquatic decomposition scoring methods

ADS methods apply numerical or categorical scoring systems to observable decomposition changes in different parts of the body and are another approach to systematically quantify and describe the postmortem changes observed in bodies recovered from water [25, 27, 29, 46, 49, 50]. These methods vary in complexity, anatomical focus, and scoring structure, but they all share the same aim of improving the consistency of PMSI estimations and the understanding of taphonomic processes in aquatic environments [25, 29, 46]. There is also emergent literature showing promise of total aquatic decomposition scores for PMSI estimations [51]. One of the first versions devised for estimating PMSI examined 187 riverine cases in the United Kingdom, using both accumulated degree days (ADD) and the scoring method of observed changes in the head, torso, and limbs [46]. Contextual modifiers included the biological profile, clothing, and deceased's body position in the water (e.g., prone, lateral recumbent, supine, etc.). Descriptive stages of observed decomposition were devised for the face, torso, and limbs, and included changes such as bloating, marbling, skin slippage, hair sloughing, adipocere formation (saponification), skeletonization, and water‐related maceration of the hands and feet [46]. This method was further refined for saltwater contexts [26] with the recovery of 184 bodies from a major shipwreck incident in the Mediterranean Sea, ranging between 3 to 15 months PMSI. The authors found that PMSI had a statistically significant relationship to summative scores for each body region and the overall body score. This version of the scoring method included more cutaneous changes, further stratification of discoloration categories, among other factors related to the appearance of the body. As a comparison, another ADS method was developed using 38 cases involving recovered persons from the North Sea [29]. When comparing these existing methods, it is clear that they were devised from open water environmental conditions that did not include anthropogenically constructed environments with treated water for comparison, thus, they do not include cases from controlled or quasi‐controlled aquatic environments. Moreover, existing ADS methods appear to suggest a linear transition between descriptive phases, and it is unclear how the intrinsic moderating factors like BMI and age, or extrinsic factors like body position and scavenging, may modify decomposition change.

Despite the significant contributions made by these and other studies to current knowledge of aquatic decomposition changes and their relationship to PMSI, further research is needed, particularly in cases of controlled or quasi‐controlled aquatic settings such as domestic anthropogenically constructed environments. Literature examining human deaths in pools, bath tubs, spas, and hot tubs has largely been limited to investigations of manner and cause of death rather than body decomposition [52], and has been presented as case studies [24, 53], and retrospective studies utilizing autopsy records [22, 31]. Knowing the length of immersion is critical for establishing a baseline of decomposition changes observed in aqueous environments, and it is also important to assess whether partial immersion, clothing, individual biometrics, and other moderating factors will meaningfully alter the relationship between PMSI and specific taphonomy. The aim of this study was to augment current understanding of which variables influence the aquatic decomposition process and thus influence the PMSI by conducting a retrospective review of aquatic fatalities in anthropogenically constructed environments. This aim was addressed by taking a quantitative approach to first assess which decomposition changes in the head, torso, and limbs were statistically related to PMSI; then to assess the effects of extrinsic (situational) and intrinsic (demographic and anatomical size) factors on PMSI; and finally, to test the viability of covariates as contributors to predictive models of taphonomic change.

2. MATERIALS AND METHODS

2.1. Data source

This study involved a retrospective case review of aquatic decomposition in domestic water environments (pools, bathtubs, spas, and hot tubs) that occurred over a 10‐year period between January 1st, 2010, and December 31st, 2019, in the state of Victoria, Australia. This process involved screening, collating, coding, and developing a secondary database on all aquatic deaths, with information derived from a combination of police, autopsy, and coroner reports, mortuary photographs, and ancillary (e.g., forensic toxicology, anthropology, and entomology) reports from the Victorian Institute of Forensic Medicine's (VIFM) integrated case management system (iCMS). Potential cases were initially identified via full‐text query (water*, aqua*, submerge*, drown*, and immersion*) of Structured Query Language (SQL). Each case was then reviewed several times to ensure that they met the inclusion criteria of age (only adults aged 16 or older because we cannot assume that children are anatomically similar to adults), relevancy, within project scope (excluding electrocutions and hot water burns), time period, context (pools, bathtubs, hot tubs, and spas) and geographic location (state of Victoria). Cases that did not meet these criteria were excluded (Figure 1).

FIGURE 1.

Flowchart depicting case inclusion and exclusion.

2.2. Measures and coding practices

For the cases that met the study inclusion criteria, the PMSI (main dependent variable), extrinsic situational factors, and intrinsic demographic and anatomical size factors were recorded (Table 1). The estimated PMSI for each case was obtained from the coronial records, which documented a variety of methods used by police and coroners to discern the time of water entry as accurately as possible, including eyewitness accounts, video footage, and phone records. Decomposition was assessed using criteria adapted from existing literature and ADS systems [16, 25, 26, 29, 50, 54] and guided by ANSI/ASB Standard 149 [48], which involved reviewing their scoring criteria, groupings, and coding practices. The Franceschetti et al. [26] system is an adjusted version of the De Donno et al. [25] ADS method, which itself was modified from Heaton et al. [46] that derived from Megyesi et al. [55], which was built on the seminal work of Galloway et al. [56]. Following a similar intellectual progeny, the van Daalen et al. [29] method was devised from, and bears resemblance to, the work of Megyesi et al. [55], despite these being two distinct scoring methods. For this study, body decomposition and the taphonomic variables were coded dichotomously for the absence and presence of taphonomic change, specific to the head (including neck), torso, and limb regions; variables included bloating (gaseous distension), saponification, and dermal travel‐induced abrasion (from suspected contact with pool surfaces). Discoloration was coded categorically as normal, pink/red, or purple/brown/black. The manner of death, perimortem activity (the activity the deceased was engaged in at the time of death), and commonly documented non‐diagnostic body‐related findings were coded and included for descriptive statistics only. Categorical variables were dichotomized per level for regression modeling. Observations of rigor and livor mortis could not be reliably discerned from the scene photographs and although these changes were noted in the pathology reports, hours to days may have elapsed between the time of recovery and the autopsy. In some cases, such changes were not accounted for because the recovered bodies had been placed into holding fridges. Other commonly reported decomposition changes of cutis anserina, corneal clouding, irides graying, and petechial hemorrhages were seen in the sample; however, they could not be analyzed due to their limited predictive value. To mitigate confirmation bias during the decomposition assessment process, the situational factors were collected separately and the order in which cases were reviewed was randomized to ensure that the reviewer was blind to the approximate PMSIs during the image coding.

TABLE 1.

Summary of variables collected, coded, and analyzed.

Category	Variables
Dependent variable	Postmortem Submergence Interval (continuous, hours)
Head aquatic decomposition (HAD)	Bloating (presence or absence), Disarticulation (presence or absence), Discoloration (normal, pink/red, purple/brown/black), Hair Avulsion (presence or absence), Saponification (presence or absence), Scalp Decomposition (presence or absence), Skeletonization (presence or absence), Travel‐Induced Abrasion (presence or absence)
Torso aquatic decomposition (TAD)	Bloating (presence or absence), Disarticulation (presence or absence), Discoloration (normal, pink/red, purple/brown/black), Marbling (presence or absence), Saponification (presence or absence), Skeletonization (presence or absence), Skin Sloughing (presence or absence), Travel‐Induced Abrasion (presence or absence)
Limbs aquatic decomposition (LAD)	Bloating (presence or absence), Disarticulation (presence or absence), Discoloration (normal, pink/red, purple/brown/black), Marbling (presence or absence), Saponification (presence or absence), Skeletonization (presence or absence), Skin Sloughing/Slippage of hands or feet (presence or absence), Travel‐Induced Abrasion (presence or absence)
Extrinsic factors (situational moderators)	Anchored Down (presence or absence), Clothing (nude, partial exposure, limited exposure), Drug Use (presence or absence), Disemboweled (presence or absence), Dismemberment (presence or absence), Immersion (partial/floating or fully submerged), Incised Wrists (presence or absence), Location (pool or bathtub/hot tub/spa), Position (prone, lateral recumbent, supine or other), Sequestered (presence or absence), Scavenged (presence or absence)
Intrinsic factors (demographic and anatomical size moderators)	Age (continuous, years), BMI (continuous), BMI Classification (ordinal), Height (cm), Weight (kg), Sex (male or female)
Non‐specific body‐related findings (not moderators)	Emphysema Aquosum (presence or absence), Pleural Effusion (presence or absence), Pulmonary Oedema (presence or absence), Blood– Auditory Canal (presence or absence), Blood – Mouth (presence or absence), Blood – Nares (presence or absence), Foam Plume (nasal and/or oral) (presence or absence), Lip Discoloration (presence or absence)
General case information	Case ID, Study ID, Manner of Death (accidental, suicide, homicide, natural, undetermined), Perimortem Activity (bathing/soaking, swimming, unintentional water entry, or other (intentional water entry))

2.3. Statistical and analytical approach

Using RStudio (v.2025.05.1 + 513) and IBM SPSS Statistics (v.30.0.0.0.172), descriptive statistics, bivariate analyses, and regression modeling were conducted to identify significant relationships between PMSI and decomposition at 95% confidence intervals (0.5 alpha) and skewness and kurtosis tolerance of ±2. The type of bivariate tests deployed to measure each interaction was directed by the data itself to improve accuracy of findings: (i) Chi‐squared (χ²), whereby symmetric measures of effect size include Phi and Cramer's V; (ii) Independent t‐tests (t), where Cohen's d was used to measure effect size, and adjusted Glass's ∆ when there were unequal variances, as signified by a significant Levene's test for the equality of variances. The effect size, as measured by Glass's ∆, is greater than one because the difference between means is more than one standard deviation of the reference group, indicating a very large effect size; (iii) ANOVA (F) was used in lieu of t‐tests in scenarios involving PMSI and a taphonomic variable with 2+ categorical levels. Epsilon‐squared (ε²) was used as the adjusted eta‐squared (η²) to account for the high degrees of freedom in the model and error term, for a more accurate estimation of the effect size. Univariate General Linear Model (GLM) (Two‐Way Factorial ANOVA) was conducted to evaluate the effect of moderating covariates when a significant relationship was found between PMSI and taphonomic factors. Each of the two types of moderators were evaluated: (i) situational factors related to the circumstances of the death (outlined above), (ii) demographic and size factors pertaining to the decedent's age, sex, height, weight, and BMI. Alone, these moderating variables cannot theoretically predict the PMSI, but it was hypothesized that they would influence the direction and/or strength of the independent variables (i.e., all of the decomposition changes observed). Partial eta squared (η² p) effect sizes for moderators had the thresholds of small (<0.06), moderate (0.06–0.14), and large effects (>0.14). Multiple linear regression was conducted to assess how well decomposition changes and moderators were able to explain the variance observed in PMSI. An adjusted R ² (coefficient of determination) score was used. All categorical variables with 2+ levels were dichotomized given that linear regression models calculate the coefficients (slopes) that represent change in the dependent variable (DV) for unit changes in the independent variable (IV). Variables with insufficient data were excluded from the final model comparisons. BMI and BMI classification were independently tested for predictive power but ultimately excluded in favor of height and weight to control for issues of multicollinearity. Given that the adjusted R ² penalizes the addition of erroneous variables, this score was used to determine which variables would ultimately be included.

3. RESULTS

3.1. Overview

There were 79 cases that met the study inclusion criteria; 35 recovered from a bathtub/hot tub/spa, and 44 recovered from a pool. The manner of death was predominantly accidental (51.9%), followed by suicide (26.6%), natural (12.7%), and undetermined (8.9%); there were no homicides (Table 2). In five of the suicides, the deceased anchored themselves down by wearing heavy backpacks or binding their neck and/or wrists to heavy objects (bricks and workout weights). Sharp force injury in the form of incised wrists was observed in six of the suicide cases, and there was one case of disembowelment. The sample population comprised 53.2% (42) males and 46.8% (37) females, ranging in age from 16 to 97 years old, in height from 139 cm to 188 cm and weight from 37 to 129 kg (Table 3). Descriptive statistics for the sample population are shown in Table 3.

TABLE 2.

Frequency distribution of aquatic cases recovered from anthropogenically constructed environments by manner of death (n = 79).

	Manner of death					Total per location
	Natural	Accident	Suicide	Homicide	Undetermined
	% (n)	% (n)	% (n)	% (n)	% (n)
Bathtub a	17.1 (6)	34.3 (12)	37.1 (13)	0 (0)	11.4 (4)	35
Pool	9.1 (4)	65.9 (29)	18.2 (8)	0 (0)	6.8 (3)	44
Total	12.7 (10)	51.9 (41)	26.6 (21)	0 (0)	8.9 (7)	79

^a The bathtub location includes spas and hot tubs.

TABLE 3.

Descriptive statistics and data availability for cases recovered from aquatic anthropogenically constructed environments (n = 79).

Variables	% (n)	% missing (n)
PMSI (hours)	$\overline{x}$ = 18.33, Mdn = 6.35, SD = 34.03	0 (0)
Decomposition
Head aquatic decomposition
Bloating (Gaseous Distension – Face)	13.9 (11)	0 (0)
Disarticulation	0 (0)	0 (0)
Discoloration	–	0 (0)
Normal	35.4 (28)	–
Pink/Red	41.8 (33)	–
Purple/Brown/Black	22.8 (18)	–
Hair Avulsion	2.5 (2)	1.3 (1)
Saponification	0 (0)	1.3 (1)
Scalp Decomposition	1.3 (1)	1.3 (1)
Skeletonization	0 (0)	0 (0)
Travel‐Induced Abrasion	0 (0)	0 (0)
Torso aquatic decomposition
Bloating (Gaseous Distension – Torso)	17.7 (14)	0 (0)
Disarticulation	0 (0)	0 (0)
Discoloration	–	0 (0)
Normal	72.2 (57)	–
Pink/Red	17.7 (14)	–
Purple/Brown/Black	10.1 (8)	–
Marbling	27.8 (22)	0 (0)
Saponification	0 (0)	1.3 (1)
Skeletonization	0 (0)	0 (0)
Skin sloughing	8.9 (7)	0 (0)
Travel‐induced abrasion	0 (0)	1.3 (1)
Limb aquatic decomposition
Bloating (gaseous distension – limbs)	0 (0)	0 (0)
Disarticulation	0 (0)	0 (0)
Discoloration	–	0 (0)
Normal	51.9 (41)	–
Pink/Red	36.7 (29)	–
Purple/Brown/Black	11.4 (9)	–
Marbling	12.7 (10)	0 (0)
Saponification	0 (0)	1.3 (1)
Skeletonization	0 (0)	0 (0)
Skin sloughing (incl. slippage hands/feet)	7.6 (6)	0 (0)
Travel‐induced abrasion	0 (0)	1.3 (1)
Situational factors (moderators)
Anchored Down	6.3 (5)	0 (0)
Clothing	–	0 (0)
Nude	34.2 (27)	–
Partial Exposure	44.3 (35)	–
Limited Exposure (fully clothed)	21.5 (17)	–
Drug Use (Toxicity Confirmed)	10.1 (8)	3.8 (3)
Disembowelment	1.3 (1)	0 (0)
Dismemberment	0 (0)	0 (0)
Immersion	–	0 (0)
Partial/Floating	73.4 (58)	–
Fully Submerged	26.6 (21)	–
Incised Wrists	7.6 (6)	3.8 (3)
Location	–	0 (0)
Bathtub/Hot Tub/Spa	44.3 (35)	–
Pool	55.7 (44)	–
Position (Found During Recovery)	–	0 (0)
Prone	49.4 (39)	–
Lateral Recumbent	15.2 (12)	–
Supine	11.4 (9)	–
Other	24.1 (19)	–
Sequestered	0 (0)	0 (0)
Scavenged	0 (0)	0 (0)
Demographic & anatomical size factors (moderators)
Age (continuous)	$\overline{x}$ = 55.78, Mdn = 56.00, SD = 21.66	0 (0)
BMI – Adult Only Calculation (16 and older)	$\overline{x}$ = 26.90, Mdn = 26.70, SD = 6.43	0 (0)
BMI Classification – Adults (16 and older)	–	0 (0)
BMI: Underweight (<18.5)	5.1 (4)	–
BMI: Normal Weight (18.5–24.9)	34.2 (27)	–
BMI: Overweight (25–29.9)	34.2 (27)	–
BMI: Obese I (25–29.9)	11.4 (9)	–
BMI: Obese II (35–39.9)	11.4 (9)	–
BMI: Obese III (>40)	3.8 (3)	–
Height (cm) (continuous)	$\overline{x}$ = 165.29, Mdn = 166.00, SD = 10.37	0 (0)
Weight (kg) (continuous)	$\overline{x}$ = 73.68, Mdn = 73.00, SD = 20.34	0 (0)
Sex	–	0 (0)
Male	53.2 (42)	–
Female	46.8 (37)	–
Non‐specific body‐related findings (non‐moderators)
Autopsy finding – Emphysema Aquosum	50.6 (40)	13.9 (11)
Autopsy finding – Pleural Effusion	31.6 (25)	13.9 (11)
Autopsy finding – Pulmonary Oedema	59.5 (47)	13.9 (11)
Blood – Auditory Canal	0 (0)	3.8 (3)
Blood – Mouth	26.6 (21)	1.3 (1)
Blood – Nares	49.4 (39)	2.5 (2)
Foam Plume (nasal and/or oral)	27.8 (22)	2.5 (2)
Lip discoloration	–	0 (0)
Normal	67.1 (53)	–
Blanched	12.7 (10)	–
Blue Tint	8.9 (7)	–
Purple/Maroon	11.4 (9)	–
General case information (non‐moderators)
Activity (Perimortem)	–	3.8 (3)
Bathing/Soaking	26.6 (21)	–
Swimming	30.4 (24)	–
Unintentional Water Entry	12.7 (10)	–
Other (intentional water entry)	26.6 (21)	–
Manner of death	–	0 (0)
Natural	12.7 (10)	–
Accidental	51.9 (41)	–
Suicide	26.6 (21)	–
Homicide	0 (0)	–
Undetermined	8.9 (7)	–

The relationship between the anthropogenically constructed environment type (pool or bathtub/hot tub/spa) and manner of death was statistically significant ($x^2$(3) = 7.859, p = 0.049 < 0.05 (95% CI), Cramer's V = 0.315). All bodies were found intact, so there were no cases of dismemberment, disarticulation, or skeletonization. There were also no instances of bodies found in sequestered environments, nor signs of vertebrate scavengers. Even among the pool deaths where there was a possibility for the body to make contact with the pool surfaces, no travel‐induced abrasions were observed and there were also no signs of saponification or gaseous distension of the limbs. All decomposition variables absent from the dataset were excluded from further analyses. Non‐diagnostic body‐related findings included foam plume in the outer nasal and oral cavities in 27.8% (22) of cases, and blood was observed in the mouth and nose, and there was lip discoloration (Table 3). PMSI ranged from near‐immediate interventions to upwards of 168 h. The average PMSI was 18.33 h (median of 6.35 h), representing a highly leptokurtic (9.699) distribution with a positive skew (3.048), thus, bivariate tests were conducted with corrected values because the normality of distributions (±2) could not be assumed.

3.2. Head aquatic decomposition (HAD)

There was a significant association between slight gaseous distension and skin discoloration in the head region and PMSI. Slight gaseous distension (i.e., bloating) of the face was observed in 11 cases and had a significant relationship with PMSI, where equal variances cannot be assumed (t (10.455) = 2.361, p = 0.039 < 0.05 (95% CI), Glass's ∆ = 1.929; Tables 3 & 4). This feature was observed in both partially and fully submerged cases from as early as four hours up until 168 h PMSI (the tail end of the dataset's PMSI spectrum). The lowest bound PMSI of a fully immersed body was 16.33 h, which suggested that this feature was only truly apparent in those bodies found with faces fully under water from 16 h onward. Drug use and immersion did not affect gaseous distension of the face, but location and position of the body had a significant strong effect (Table 4). Among the different demographic factors, only BMI and height of the individual influenced gaseous distension of the face (Table 5). Skin discoloration of the head also showed a significant relationship with PMSI (Table 4), with the mean normal coloration of 8.2 h PMSI, mean pink/red discoloration of 12.7 h PMSI, and mean purple/brown discoloration of 44.3 h PMSI (ANOVA F (2) = 8.179, p = <0.001 < 0.05 (95% CI), ε ² = 0.155). The important finding was not that there was discoloration, but instead, the absence of it in cases with long immersion periods. One case, involving a body that was fully immersed (at time of recovery) in a pool setting with an estimated PMSI of 121.17 h, had relatively normal skin appearance. Whether the body was partially or fully immersed was found to be a significant modifier of discoloration, but the effect size was only moderate (Table 4). When it came to discoloration, the decedent's age, BMI, weight, and sex were all found to be significant moderators, though only age, BMI and weight had noteworthy effect sizes (Table 5). The remaining head taphonomic changes were not shown to be significant. Hair avulsion was only present in three cases, one of which also showed signs of scalp decomposition. There were no signs of saponification or travel‐induced abrasions.

TABLE 4.

Bivariate associations between key decomposition variables and PMSI in cases recovered from aquatic anthropogenically constructed environments (pools and bathtubs/hot tubs/spas) while testing the effects of moderating situational factors (n = 79).

Main interaction		Moderators
		Univariate GLM (two‐way factorial ANOVA) of taphonomic variables vs. PMSI (hours), controlling for moderating covariates
Variables	PMSI (hours)	Clothing	Drug use	Immersion	Incised wrists	Location	Position
	n (p) [effect size]	F (p) η2p	F (p) η2p	F (p) η2p	F (p) η2p	F (p) η2p	F (p) η2p
HAD
Bloating	79 (0.039) a , c [1.929]	– Not applicable	0.179 (0.675) NS	1.988 (0.163) NS	– Not applicable	6.737 (0.011) 0.082	6.732 (<0.001) 0.221
Disarticulation	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size
Discoloration	79 (<0.001) b [0.155]	– Not applicable	1.743 (0.183) NS	3.741 (0.028) 0.093	– Not applicable	1.391 (0.255) NS	1.412 (0.231) NS
Hair Avulsion	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size
Saponification	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size
Scalp Decomp	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size
Skeletonization	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size
TI‐Abrasion	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size
TAD
Bloating	79 (0.005) a , c [2.821]	4.040 (0.048) 0.052	0.182 (0.671) NS	1.540 (0.218) NS	– Not applicable	0.145 (0.704) NS	0.934 (0.429) NS
Disarticulation	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size
Discoloration	79 (<0.001) b [0.290]	2.345 (0.080) NS	0.060 (0.808) NS	7.943 (<0.001) 0.179	– Not applicable	3.301 (0.042) 0.083	2.620 (0.032) 0.162
Marbling	79 (0.004) a , c [1.842]	3.802 (0.027) 0.094	0.569 (0.453) NS	0.124 (0.726) NS	– Not applicable	1.516 (0.222) NS	0.403 (0.751) NS
Saponification	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size
Skeletonization	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size
Skin Sloughing	79 (0.014) a , c [4.173]	9.457 (0.003) 0.113	– Insufficient size	0.002 (0.963) NS	– Not applicable	0.280 (0.598) NS	4.396 (0.007) 0.157
TI‐Abrasion	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size
LAD
Bloating	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Disarticulation	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Discoloration	79 (<0.001) b [0.194]	– Not applicable	0.064 (0.800) NS	3.547 (0.034) 0.089	0.143 (0.867) NS	2.028 (0.139) NS	2.605 (0.025) 0.189
Marbling	79 (0.012) a , c [3.363]	– Not applicable	0.438 (0.510) NS	2.125 (0.149) NS	0.059 (0.808) NS	21.703 (<0.001) 0.224	11.245 (<0.001) 0.322
Saponification	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Skeletonization	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Skin Sloughing	79 (0.004) a , c [5.527]	– Not applicable	0.661 (0.419) NS	0.610 (0.437) NS	4.499 (0.037) 0.059	2.905 (0.092) NS	5.023 (0.003) 0.175
TI‐Abrasion	– Insufficient size	– Not applicable	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size

^a Independent t‐test (t) was used to determine the relationship significance between the dichotomous taphonomic factor and PMSI as a continuous measure of time elapsed.

^b An ANOVA (F) statistic was used to determine the relationship significance between the taphonomic factor with 2+ categories and the PMSI; Epsilon‐Squared was the measure of effect size (corrects for positive bias in eta‐squared).

^c Levene's test was significant, thus, the adjusted scores were used where equal variances could not be assumed; Glass's delta was the measure of effect size (uses the sample SD of the control).

TABLE 5.

Bivariate associations between key decomposition variables and PMSI in cases recovered from aquatic anthropogenically constructed environments (pools and bathtubs/hot tubs/spas) while testing the effects of moderating demographic and anatomical size factors (n = 79).

Main interaction		Moderators
		Univariate GLM (two‐way factorial ANOVA) of taphonomic variables vs. PMSI (hours), controlling for moderating covariates
Variables	PMSI (hours)	Age	BMI	BMI classifications	Height	Weight	Sex
	n (p) [E.S.]	F (p) η2p	F (p) η2p	F (p) η2p	F (p) η2p	F (p) η2p	F (p) η2p
HAD
Bloating	79 (0.039) a , c [1.929]	0.490 (0.693) NS	13.852 (0.002) 0.480	1.959 (0.111) NS	8.895 (<0.001) 0.634	2.137 (0.097) NS	3.969 (0.050) NS
Disarticulation	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Discoloration	79 (<0.001) b [0.155]	4.458 (0.012) 0.853	20.113 (0.002) 0.976	0.808 (0.622) NS	1.639 (0.133) NS	2.755 (0.033) 0.688	0.644 (0.528) NS
Hair Avulsion	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Saponification	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Scalp Decomp	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Skeletonization	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
TI‐Abrasion	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
TAD
Bloating	79 (0.005) a , c [2.821]	1.257 (0.323) NS	4.276 (0.026) 0.497	2.761 (0.034) 0.140	16.933 (<0.001) 0.795	6.169 (<0.001) 0.712	13.676 (<0.001) 0.154
Disarticulation	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Discoloration	79 (<0.001) b [0.290]	1.017 (0.464) NS	90.114 (<0.001) 0.970	1.161 (0.338) NS	1.822 (0.085) NS	14.556 (<0.001) 0.909	4.070 (0.021) 0.100
Marbling	79 (0.004) a , c [1.842]	0.850 (0.574) NS	58.504 (<0.001) 0.978	2.394 (0.047) 0.152	9.106 (<0.001) 0.798	4.791 (0.002) 0.756	2.024 (0.159) NS
Saponification	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Skeletonization	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Skin Sloughing	79 (0.014) a , c [4.173]	1.017 (0.498) NS	0.700 (0.513) NS	17.183 (<0.001) 0.428	15.907 (<0.001) 0.721	8.151 (<0.001) 0.494	17.808 (<0.001) 0.192
TI‐Abrasion	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
LAD
Bloating	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Disarticulation	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Discoloration	79 (<0.001) b [0.194]	0.851 (0.602) NS	32.613 (<0.001) 0.980	3.339 (0.004) 0.268	1.154 (0.361) NS	5.116 (0.003) 0.846	0.386 (0.681) NS
Marbling	79 (0.012) a , c [3.363]	6.263 (0.007) 0.363	71.080 (<0.001) 0.960	8.415 (<0.001) 0.331	16.483 (<0.001) 0.790	8.745 (<0.001) 0.512	7.862 (0.006) 0.095
Saponification	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Skeletonization	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size
Skin Sloughing	79 (0.004) a , c [5.527]	0.597 (0.940) NS	0.700 (0.513) NS	8.012 (<0.001) 0.186	11.528 (<0.001) 0.542	7.763 (0.010) 0.223	7.971 (0.006) 0.096
TI‐Abrasion	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size	– Insufficient size

^a Independent t‐test (t) was used to determine the relationship significance between the dichotomous taphonomic factor and PMSI, as a continuous measure of time elapsed.

^b An ANOVA (F) statistic was used to determine the relationship significance between the taphonomic factor with 2+ categories and the PMSI.

^c Levene's test was significant; thus, the adjusted scores were used where equal variances could not be assumed.

3.3. Torso aquatic decomposition (TAD)

A significant association was found between PMSI and bloating of the torso, discoloration marbling, and skin sloughing. Gaseous distension of the torso was observed in 16.7% (14) of cases, with the earliest appearance at 11 h PMSI, but primarily present in cases involving more than 24 h PMSI (Table 3). Its relationship with PMSI was statistically significant with a strong effect size (t (14) = 3.377, p = 0.005 < 0.05 (95% CI), Glass's ∆ = 2.821; Table 4). Clothing and sex of the individual had small and moderate effect sizes, respectively, as moderators of PMSI and torso bloating (Table 4). The individual's size (height, weight, and BMI) had a much stronger effect on PMSI and torso bloating (Table 5). Nearly three quarters (72.9%, 62) of the cases did not exhibit any discoloration of the torso even at the upper end of PMSI (Table 3). Of the remainder that did have discoloration, there were 14 cases of pink/red discoloration (16.5%) and eight cases of purple/brown/black discoloration (9.5%). The spectrum of color was difficult to discern due to variable photo quality so these were coded conservatively and similar colors were grouped together. These three categories of discoloration were found to be significantly related to PMSI (ANOVA F (2) = 16.959, p = <0.001 < 0.05 (95% CI), ε ² = 0.290), with immersion, location, position of the body, and sex found to be minor to moderate moderators (Table 4). BMI, and particularly weight, were considered much stronger and significant moderators of the relationship of discoloration to PMSI (Table 5). Marbling of the torso was observed in 28.2% (24) of cases and was significantly related to PMSI (t (23) = 3.178, p = 0.004 < 0.05 (95% CI), Glass's ∆ = 1.842). Clothing was the only situational factor that had an influence on this relationship, albeit with only a moderate effect (Table 4). However, the size of the individual (BMI, height, and weight) was found to be a very strong moderator (Table 5). Similarly, the relationship between skin sloughing of the torso and PMSI was significant (t (6) = 3.427, p = 0.014 < 0.05 (95% CI), Glass's ∆ = 4.173) and occurred as early as 11 h PMSI. The presence of clothing and position of the body, BMI, height and weight all had a significant moderate effect on skin sloughing of the torso (Tables 4 and 5). Partial immersion resulted in localized discoloration, whereby the portion of the body immersed in water remained blanched and the portion above water underwent discoloration; this was more evident in bodies with longer PMSI.

3.4. Limbs aquatic decomposition (LAD)

PMSI was significantly associated with discoloration, marbling, and skin sloughing of the limbs. Discoloration of the limbs was observed in 45.9% of cases, whereby 35.3% (30) had pink/red discoloration, and 10.6% (9) had purple/brown/black discoloration. The relationship between discoloration of limbs and PMSI was found to be statistically significant, and the appearance of any discoloration ranged from 6–168 h PMSI (ANOVA F (2) = 10.396, p = <0.001 < 0.05 (95% CI), ε ² = 0.194). Immersion had a moderate effect on the relationship between discoloration and PMSI, and the position of the body, BMI, and weight were all found to have stronger effects (Tables 4 and 5). Marbling and PMSI had a significant relationship, with this feature of decomposition found to be present in 14.1% (12) of cases (t (9) = 3.123, p = 0.012 < 0.05 (95% CI), Glass's ∆ = 3.363) and moderated by location and position of the body (Table 4). With respect to demographic and size factors, the decedent's age, BMI, height, and weight all had strong effects on this relationship (Table 5). Skin sloughing of the limbs was also significantly associated with PMSI (t (5) = 5.116, p = 0.004 < 0.05 (95% CI), Glass's ∆ = 5.527). In cases of suicide by incision of the wrists, exsanguination had only a weak influence on the relationship between PMSI and skin sloughing (Table 4); this may be due to the rapid recovery of those individuals, so the impact of blood loss on decomposition could not be fully tested. The position of the body had a greater impact on the relationship between PMSI and skin sloughing, as did BMI group, height, and weight (Tables 4 and 5).

3.5. Regression models

Five models were generated to demonstrate which variables were significant contributors to model creation when covariates were considered. The key taphonomic variables from each of the three regions of the body were included in Models 1–4, alongside the strongest situational and demographic/size moderators, as identified in the bivariate analyses (Table 5). Because adjusted R² values correct for auto‐inflation, such models penalize the inclusion of more variables, so every newly added variable must be useful enough to outpace the drop in the model's predictive value; adjusted scores will optimize for parsimony, whereby it will penalize models for every additional variable included. Compared to Model 1, which only considered decomposition variables, the inclusion of only situational moderators (Model 2) or only demographic/size moderators (Model 4) did not improve the overall predictive value enough to offset the penalty (Table 6). Inclusion of all moderators (Model 3) resulted in a 1.8% drop (0.595 to 0.577), and the standard error increased. Overall, Models 1–4 all performed similarly (within 2% difference) when all decomposition variables were considered.

TABLE 6.

Multiple linear regression modeling of PMSI, decomposition variables, and moderating factors in cases recovered from aquatic anthropogenically constructed environments (pools and bathtubs/hot tubs/spas) (n = 79).

Variables	Model 1 (Decomp only)	Model 2 (+ SF)	Model 3 (+ SF & DEM)	Model 4 (+ DEM)	Model 5 (selective)
	b (SE) [p]	b (SE) [p]	b (SE) [p]	b (SE) [p]	b (SE) [p]
Decomposition
HAD – Bloating	−3.829 (9.773) [0.696]	0.070 (10.945) [0.995]	−2.447 (11.999) [0.839]	−6.785 (10.054) [0.502]	–
HAD – Discoloration a	2.243 (5.873) [0.704]	−1.260 (6.381) [0.844]	−2.622 (6.646) [0.695]	−0.158 (6.142) [0.980]	–
TAD – Bloating	2.440 (12.638) [0.847]	2.370 (12.844) [0.854]	6.541 (14.188) [0.646]	8.587 (13.526) [0.528]	2.446 (9.294) [0.793]
TAD – Discoloration a	6.393 (9.095) [0.484]	6.995 (9.326) [0.456]	7.119 (9.546) [0.459]	6.784 (9.200) [0.464]	–
TAD – Marbling	4.890 (9.090) [0.592]	−0.349 (10.257) [0.973]	−0.980 (11.047) [0.930]	3.505 (9.455) [0.712]	–
TAD – Skin Sloughing	18.518 (15.381) [0.233]	15.821 (15.750) [0.319]	15.939 (16.523) [0.338]	17.173 (15.940) [0.285]	19.536 (14.034) [0.168]
LAD – Discoloration a	−5.399 (6.197) [0.387]	−5.149 (6.257) [0.413]	−6.658 (6.519) [0.311]	−6.636 (6.403) [0.304]	–
LAD – Marbling	−4.269 (12.778) [0.739]	1.020 (13.404) [0.940]	2.033 (13.718) [0.883]	−1.380 (13.097) [0.916]	–
LAD – Skin Sloughing/Slippage	82.978 (17.581) [<0.000]***	76.855 (18.156) [<0.000]***	76.429 (18.563) [<0.000]***	80.782 (17.876) [<0.000]***	78.390 (13.977) [<0.000]***
Situational Factors
Clothing – Naked b		0.746 (7.202) [0.918]	0.736 (7.523) [0.922]	–	–
Immersion		5.108 (6.194) [0.413]	4.346 (6.876) [0.530]	–	–
Location – Bathtub c		8.436 (7.962) [0.293]	7.326 (9.417) [0.440]	–	9.115 (5.446) [0.099]+
Demographic Factors
Age			0.042 (0.138) [0.761]	−0.008 (0.125) [0.950]	–
Height (cm)			0.232 (0.368) [0.530]	0.211 (0.355) [0.555]	0.069 (0.289) [0.812]
Weight (kg)			0.089 (0.159) [0.579]	0.085 (0.156) [0.590]	0.055 (0.145) [0.706]
Sex			8.439 (7.198) [0.246]	10.506 (6.802) [0.127]	–
Adjusted R 2	0.595	0.591	0.577	0.588	0.618
SE of Estimates	21.64733	21.76425	22.14676	21.85287	21.03069
Constant b (SE)	9.030 (4.354)	6.617 (4.725)	42.697 (61.038)	−34.865 (58.790)	−0.158 (42.149)
Akaike Information Criterion	495.139	498.479	504.292	499.915	487.936
Amemiya Prediction Criterion	0.462	0.482	0.521	0.492	0.421
Mallows' Prediction Criterion	6.923	10.740	17.000	14.000	7.000
Schwarz Bayesian Criterion	518.834	529.282	544.573	533.087	504.522

Note: + p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.001.

^a All categorical discoloration decomposition variables were dichotomized for linear regression modeling (0 = normal, 1 = change observed).

^b Clothing presence was dichotomized (0 = clothed, 1 = naked).

^c Location reclassified (0 = pool, 1 = bathtub/hot tub/spa).

To improve model accuracy, more than 30 alternative stepwise comparisons with all available variables were computed to discern the version with the best adjusted R² and lowest standard error. The result was several models that excluded variables pertaining to decomposition of the head altogether, indicating that changes in the head and neck region were less reliable indicators of PMSI than changes in the torso or limbs. Moreover, the only factors included in the best version (Model 5) further excluded discoloration and marbling from all areas of the body and focused on bloating of the torso and skin sloughing of the torso and limbs (Table 6). Skin sloughing of the limbs and slippage of hands and feet were significant contributors across all models, so its selection for Model 5 was not surprising (Table 6). Interestingly, Model 5 reintroduced the location, height and weight as key moderators, which added support for the inclusion of specific moderators when trying to better understand PMSI. Location being a bathtub was the only other significant contributor to overall model predictive power. Overall, the best model that could be constructed with these data was Model 5, whereby 61.8% of the variance observed in the PMSI (adjusted R ² = 0.618) could be explained. Even though there were many theoretically useful variables that could not be collected, the particular selection within this dataset could account for nearly two thirds of the contributing factors that influenced PMSI. The standard error of estimates also reduced from 21.64733 (Model 1) to 21.03069 (Model 5), which was indicative of a 0.61664 improvement in the predictive accuracy, indicating Model 5 had a marginal increase in reliability.

4. DISCUSSION

The focus of this study was to improve understanding of human decomposition in quasi‐controlled, aquatic, anthropogenically constructed environments. The novelty of this work is that it represents one of the first detailed examinations of decomposition in this type of environment with a relatively large sample size with near complete taphonomic information and approximately known PMSI. Further, bodies recovered found in a lateral recumbent position in bathtubs allowed for closer examination of immersion effects because there were clearly demarcated water boundaries and stable conditions necessary to compare portions of the body found immersed with those above water. Given that the Australian Life Saving Victoria 2024–2025 Drowning Report identified that the home accounted for 25% of fatal drowning locations [57], this retrospective review offers a close approximation to real world conditions, and allows researchers to develop a baseline of what decomposition changes may be expected across three anatomical regions (head, torso, and limbs) for bodies in anthropogenically constructed environments and, most importantly, which taphonomic changes are more strongly associated with PMSI. This study also addresses the need for more robust statistical approaches that account for situations where equal variances cannot be assumed. Results of main bivariate associations between PMSI and decomposition variables also directed the use of a Univariate GLM, which allowed for the quantification of moderating effects from situational, demographic, and size factors that are commonly cited in the literature but rarely tested. Moreover, these specific approaches challenged the idea that all decomposition variables across the body are stable predictors of PMSI and offered new insights as to the importance of discerning the location, height and weight as important moderators for the onset of decomposition when other covariates were held constant. In this study, the data guided the analysis, which enabled the development of a more accurate baseline understanding of PMSI, decomposition, and effects of moderators on decomposition in aqueous pool and bathtub environments.

When examining aquatic decomposition of the head region (HAD), bloating and discoloration were both individually statistically significantly related to PMSI but were found to be poor contributors to overall model prediction of variances observed in PMSI. These findings highlight the importance of going beyond bivariate analyses in assessing the presence of taphonomic variables, because the best regression model produced in this study showed that HAD introduced more noise and erroneous variance than it contributed to the overall predictive power of the model. Regarding aquatic decomposition of the torso (TAD), bloating, marbling, and skin sloughing were all found to be significantly related to PMSI, which supports the existing literature that has taken note of such taphonomic changes [16, 32, 58]. Univariate GLM results confirmed that clothing had a minor moderating effect. For aquatic decomposition of the limbs (LAD) there were no signs of bloating. Discoloration, marbling, and skin sloughing were all found to be statistically related to PMSI at the individual, bivariate level. When considered with covariates, skin sloughing and skin slippage of the hands and feet were the only variables to retain their status as strong contributors to the overall model. Although this study can only speak to observations of water‐related deaths in anthropogenically constructed environments and not naturally occurring waters, these results highlight the importance of heightened care to safely recover hands and feet, a concern that exists particularly in lotic or oceanic environments where there are currents and water movement [58].

The anthropometric measurements of the body (height, weight, and BMI) were considerably more useful moderators of postmortem change than situational factors. BMI has already been noted in the literature for its importance [22], and the findings from this study further support this claim. The use of BMI classifications was also tested to see if this would offer better demarcated groups for analysis, but this was not the case. Retaining the raw BMI score offered stronger partial eta‐squared scores, which suggested that the variance in scores was important and would be lost when the data resolution is reduced in favor of BMI classification groups. Given these findings, one of the key recommendations from this study is for practitioners to retain the height, weight, and subsequently the linear BMI score (as a measure of height and weight combined, BMI = weight (kg) / height (m²)). It is also the case that BMI classification standards may differ between countries, so the use of BMI classifications in lieu of the BMI score may actually present as an obstacle for cross‐national comparisons. There was also evidence to suggest that age was not an important moderator, and sex had moderate effect sizes only on the torso; thus, it should not be assumed to influence decomposition in other parts of the body. Drug toxicity was not found to moderate taphonomic change, despite being present in nearly one tenth of the sample population. This result differed from that found in another retrospective review [22], but this may be due to the higher rates of toxicity found overall in their dataset.

Non‐specific features were also examined. The presence of blood‐stained fluid in the auditory canal, mouth, and nares, as well as lip discoloration and foam plume were present, but were considered non‐informative because they may also appear in non‐aqueous contexts [58], despite their common notation in water‐related deaths [16, 23, 58]. Similarly, ocular changes, including irides graying, sunken eyes, and petechial hemorrhages, which have been noted in the aquatic decomposition literature [58] and appeared commonly, offered no further benefit to understanding PMSI. It would be helpful for researchers and first responders to know that these features occur but may have been caused by other medical factors or trauma to the eyes and are not useful considerations for estimating PMSI. Autopsy findings did include emphysema aquosum, pleural effusion, and pulmonary oedema, but these were excluded from further analysis because their function as moderators of decomposition requires further investigation. Perimortem activity (i.e., what the deceased was doing at the time of death) was also considered, but was found to have a non‐significant impact, and even as a covariate, was not useful for understanding the onset of decomposition changes and PMSI, therefore there was no indication that this would be a useful moderator in pool or bathtub/hot tub/spa environments. Because these changes are not helpful towards the prediction of PMSI, they should be considered only insofar as simply noting they may appear in water‐related cases, among other possibilities.

Methodologically, the choice to report an adjusted R² instead of R² was made intentionally to limit the effects of overfitting because this statistic penalizes the addition of irrelevant predictors to provide a more accurate measure of model fit. At present, it was not intended that the model be used to predict PMSI in casework due to the limited nature of the dataset. It is hoped that this method will be tested in the future across various contexts and with increasingly robust datasets to further improve what is known about PMSI, decomposition, and the effects of moderators. Bodies were scored using criteria adapted largely from the ADS systems described by Franceschetti et al. [26] and van Daalen et al. [29], but these methods differed substantially when applied to bodies recovered from aquatic anthropogenically constructed environments, and there were instances where the ADS order of changes did not match what was observed in the present study. This difference was not surprising given that these updated ADS methods were designed using saltwater cases and confirms that such ADS instruments should not be generalized to cases in aquatic treated freshwater constructed environments consisting of pools, bathtubs, spas, and hot tubs. More work should be completed to properly test agreement levels between the many ADS methods in existence [6, 12, 17, 25, 26, 29, 38, 39, 40, 46, 49, 55, 56, 59, 60, 61, 62, 63], across different environmental contexts. In the present study, discoloration in multiple regions of the body and other observed changes could not be captured by the van Daalen et al. [29] ADS method because its categorical scales could not account for any darkening of the body; in fact, the van Daalen et al. [29] method only mentions skin becoming a gray, matte color with a crumbly appearance, which was not observed in any of the present cases. Another example of expected taphonomic changes that were not observed in our study include hair avulsion and brain liquification, which is presented as a stage of decomposition in the Franceschetti et al. [26] scoring method; the absence of such taphonomic changes may be a result of the temporal and geographical period examined. These findings suggested that existing scoring systems devised for naturally occurring aquatic contexts may have limited application to domestic aquatic settings because not all of the same types of decomposition changes are observed and not necessarily in the same order. Additionally, anthropogenically constructed environments may contain potable and/or chemically treated water, which differs fundamentally from natural aquatic systems. The focus on early stages of PMSI in aquatic anthropogenically constructed environments also lends itself to a different temporal focus. Future research should consider developing a new scoring method that is location‐specific for bodies recovered from different aquatic contexts.

Body position was found to be an important moderator for bloating of the face, discoloration of the torso and limbs, and was also found to influence limb marbling and skin sloughing. This variation in decomposition presentation due to body position is not only relevant for bodies found in a left or right lateral recumbent position in domestic aquatic contexts; there may be scenarios, particularly in natural aquatic environments, where the body is snagged or sequestered, whereby the difference is observed between two variably sized portions of the body on any plane. It is therefore recommended that aquatic scoring methods should consider scoring both sides of the body separately to account for partial immersion effects, given that different areas of the body might be submerged for varying amounts of time. These results support literature describing that aqueous environments may slow decomposition, likely due to suspected effects of relatively cooler water compared to ambient temperatures [16]. Further work must be done to help disentangle issues of multicollinearity between variables, as in the case of immersion and location. Of note, Australian pools are sometimes filled with saltwater, which may have contributed to the observed difference between pools and bathtub/hot tub/spa. With more data, particularly with information from cases with delayed discovery, it can be surmised that these models would become more reliable. There were also no cases of dismemberment in the present dataset; therefore, future studies should test whether that may exacerbate the onset of decomposition changes. Individuals who exsanguinated were recovered quickly so it is also not known if significant blood loss would result in different patterns of decomposition. Further examination of environmental conditions may yield more reliable information, particularly when it relates to the water temperature, microbiology, and water chemistry, given recent advances in aquatic decomposition modeling [11, 12, 17, 38].

Although the bivariate results and regression model were able to offer some indication of the usefulness of moderators, it is important to note that these models have limitations. Firstly, these models have not been blind tested against other datasets and were built on cases solely from domestic aquatic settings. Statistical models could be improved with larger sample sizes, particularly with more information from cases with extended PMSI exceeding 168 h. They cannot account for other immersion contexts, such as sequestered locations or naturally occurring environmental contexts like rivers, lakes, and oceans. Secondly, due to ethical limitations, it was not possible to test interobserver error in the coding of mortuary photographs. Thirdly, scene photographs, particularly those from the years 2010–2012, varied in resolution and picture quality. Great caution was exercised by not flagging decomposition unless it appeared clear, which may have resulted in false negatives or a greater number of missing cell counts. It was believed that such a conservative approach was more appropriate than overestimating these effects. Moreover, a conservative approach was also taken with coding discoloration. It was acknowledged that pink discoloration could very well be hypostasis and red discoloration from either putrefaction or hypostasis, but the two could not confidently be discerned due to photo quality, which is why pink and red discoloration of the skin were amalgamated. Without direct contact with the body upon recovery, this is the best that could be achieved in a retrospective review. Fourthly, variables that were not reported in the case files, such as water temperature, could not be included.

At present, there is no evidence to suggest there are clear demarcated stages of decomposition in aquatic anthropogenically constructed environments. In fact, the lack of these clear stages was one of the key benefits of conducting this baseline study. The results of this research demonstrate that there are extrinsic factors that influence the rate and appearance of taphonomic change, but not uniformly, and it is also evident that the effects of immersion may actually delay or buffer against some taphonomic changes altogether, adding greater complexity to interpretations of PMSI. This study also demonstrated a statistical approach to quantifying taphonomic change that can scale as the dataset improves, allowing for growth and further refinement. Further research is necessary to improve these models and to strengthen what is known about aquatic decomposition patterns for medicolegal investigations, but this work represents a necessary step towards the goals of estimating PMSI and understanding the postmortem journey of bodies recovered from aqueous anthropogenically constructed environments.

FUNDING INFORMATION

The authors would like to thank the following funding agencies, including the American Academy of Forensic Sciences Pathology/Biology Award (Anderson), Mitacs Globalink Research Award (Lam: IT33184), Social Sciences and Humanities Research Council (SSHRC) Joseph Armand Bombardier Canadian Graduate Student Doctoral Award and SSHRC Michael Smith Foreign Study Supplement (Lam: 767‐2018‐2671), P.E.O. Scholar Award, Simon Fraser University (SFU) School of Criminology Travel and Research Award and Doctoral Graduate Fellowship, and SFU Faculty of Arts and Social Sciences Dean's Graduate Student Support Fund.

CONFLICT OF INTEREST STATEMENT

The authors do not have any conflicts of interest to declare.

ETHICS STATEMENT

This international research collaboration has received approval from both Simon Fraser University's Research Ethics Board (19s0259) and the Victorian Institute of Forensic Medicine's Ethics Committee (2022‐Lam‐1238‐2354/1).

Lam VC, Anderson GS, Archer MS, Dickson GC, Rowbotham SK. Postmortem submergence interval (PMSI) and human decomposition in anthropogenically constructed aqueous environments (pools, bathtubs, hot tubs, and spas). J Forensic Sci. 2026;71:1718–1734. 10.1111/1556-4029.70316

DATA AVAILABILITY STATEMENT

Research data are not shared.