Mild vitamin C deficiency is common in the inpatient psychiatric setting

Abstract

Objective:

Mild vitamin C deficiency is a psychiatrically-relevant nutritional state, with symptoms including apathy, fatigue, and low mood. Although complete vitamin C deficiency has largely been eradicated, mild deficiency remains common in certain populations. Here, we aimed to identify the prevalence of mild vitamin C deficiency in the inpatient psychiatric setting.

Methods:

We identified 221 patients with serum vitamin C levels collected on an inpatient psychiatric unit serving a metropolitan area between January 1, 2015 and March 7, 2022. We identified demographic (age, sex, race, housing status, Area Deprivation Index [an index of neighborhood disadvantage]), substance use (tobacco use, alcohol use), diagnostic (depressive, bipolar, psychotic, anxiety, substance use, catatonia, neurocognitive, autism spectrum), and micronutrient (folate, vitamin B12, vitamin D) risk factors. We used the DSM-5-TR as our diagnostic system. We constructed Bayesian log-normal regressions to predict vitamin C as a function of these risk factors. We used these same models to predict vitamin C as a function of significant risk factors.

Results:

We found that 64% (141 of 221; 95% confidence interval 57–70%) of patients met criteria for mild vitamin C deficiency. While we did not identify robust demographic, substance use, or diagnostic-based risk factors, we found that folate and vitamin D strongly predicted vitamin C levels. To test the utility of these predictors, we simulated vitamin C as a function of folate and vitamin D and found that predicted deficiency remained high (~50–55%), even when folate/vitamin D were sufficiently replete.

Conclusion:

We find that vitamin C deficiency is highly prevalent in the inpatient psychiatric setting and remains high even when the relevant risk factor profile is favorable.

Introduction

Scurvy is among humanity’s oldest documented conditions. First reported in the Egyptian Ebers Papyrus in 1550 BC [1, 2], scurvy is characterized by petechiae, perifollicular hemorrhage, ecchymoses, gingival bleeding, and ultimately death due to infection or bleeding [3]. Scurvy is a consequence of profound vitamin C deficiency [4, 5] and is reversed with a diet rich in citrus fruits, a discovery credited to the Scottish surgeon James Lind [6] (though see [7, 8], as Vasco da Gama may have preempted this discovery by several centuries). With current dietary recommendations [9] and modern diets [10], the burden of this once commonplace disease has been significantly diminished.

Physical symptoms, however, are not the only consequence of vitamin C deficiency. One of the earliest writings on the psychiatric consequences came from English military surgeon John Woodall, who wrote of a “generall lazinesse” in those with scurvy [11]. Lind, in his treatise, wrote “...the person eats and drinks heartily, and seems in perfect health; except that his countenance and lazy inactive disposition may portend an approaching scurvy [6].” The British physician Thomas Shapter, best known for his account of cholera, recognized a similar pattern. He wrote of the “phases” of vitamin C deficiency and documented weakness, listlessness, and a lack of motivation as symptoms of the first phase, well before the devastating symptoms of scurvy took hold [12]. Modern experiments have replicated these observations: in healthy volunteers deprived of vitamin C, the symptoms of low mood, weakness, listlessness, and lack of motivation emerge, and recover with repletion [13, 14]. We now understand these symptoms as a consequence of mild vitamin C deficiency [15, 16], a psychiatrically-relevant state of nutritional deficiency recognized by the World Health Organization and defined by a serum vitamin C level of < 0.75 mg/dl [3].

Mild vitamin C deficiency is remarkably common, and is seen in a number of geographic contexts [17] (including the United States [18, 19], France [20, 21, 22], the United Kingdom [23, 24], Canada [25, 26, 27], Australia [28], Brazil [29, 30], Mexico [31, 32], Ecuador [33], China [34, 35], Thailand [36], India [37], Nigeria [38], and Uganda [39]), medical contexts (medical populations [26, 27], and populations with psychiatric illness [28]; inpatient settings [22, 23, 25, 26, 27], and outpatient settings [18]), and socioeconomic contexts (undomiciled populations [21], and the general population [19]). Risk factors include hospitalization [16, 25], male sex [19, 22], diabetes [18], poor dietary intake [19, 24], smoking [19, 24], and low socioeconomic status [19, 21].

Vitamin C deficiency is particularly pronounced in the hospital setting, with over half of medical inpatients meeting criteria for mild vitamin C deficiency [25, 26, 27]. However, it is not well understood if vitamin C deficiency is as prevalent in the psychiatric inpatient setting. Here, we sought to identify the prevalence of vitamin C deficiency in psychiatrically-hospitalized patients. This is an important and timely question because vitamin C deficiency has known psychiatric consequences, yet is not part of the standard screening and workup for psychiatric complaints (e.g., thyroid stimulating hormone, vitamin D, vitamin B12, folate). This is compounded by the logistical challenge of obtaining vitamin C levels, since assays are typically performed at external facilities which leads to a several-day delay. If we found that vitamin C deficiency was common, then deficiency could either be treated empirically, or obtaining vitamin C levels could become a routine part of practice. We found that a substantial fraction of patients met criteria for mild vitamin C deficiency. This condition was common across our patient population and cut across all identifiable patient characteristics, as we did not establish any demographic, substance use, or diagnosis-based risk factors. We found that other markers of micronutrient status (vitamin D and folate, but not vitamin B12), were significantly correlated with vitamin C status. Taken together, we find that mild vitamin C deficiency is remarkably common in an inpatient psychiatric setting, with implications for identification and treatment.

Methods

Data collection

We obtained approval from the Massachusetts General Brigham Institutional Review Board prior to conducting this research. We identified 221 patients with serum vitamin C levels collected on the Inpatient Psychiatric Service at Massachusetts General Hospital between January 1, 2015 and March 7, 2022. Mild vitamin C deficiency was defined by serum vitamin C < 0.75 mg/dl following WHO criteria [3]. Frank deficiency, defined by < 0.2 mg/dl was found in 6% (14 of 221) of subjects. We included data from all subjects that met criteria for mild deficiency. We obtained broadly similar results when we excluded those with frank deficiency (e.g., similar conclusions, with slightly different p-values and Bayes factors). Tobacco use was defined by current smoking status as of the admission. Alcohol use was defined by an AUDIT-C [40] ≥ 4 for males and ≥ 3 for females. As a proxy for the socioeconomic status, we used the Area Deprivation Index, an index of neighborhood disadvantage [41, 42]. The Area Deprivation Index was developed and validated from Census Bureau data to provide an index relevant for healthcare outcomes [43, 44]. Area Deprivation Index spans in percentiles from 1 (lowest level of “disadvantage”) to 100 (highest level of “disadvantage”). We used the 2019 data, and linked each patient to an Area Deprivation Index using their home ZIP code.

To obtain patient diagnoses, we manually read discharge summaries and categorized admissions into at least one of eight DSM-5-TR diagnostic categories [45], allowing patients multiple diagnoses if appropriate: depressive, bipolar, psychotic, anxiety, substance use, catatonia, neurocognitive, and autism spectrum.

To assess other markers of micronutrient status, we analyzed folate, vitamin B12, and vitamin D, as these are routine admission labs. Of the 221 patients with serum vitamin C levels drawn, folate was checked in 164, vitamin B12 was checked in 181, and vitamin D was checked in 186.

Bayesian log-normal regression

We predicted vitamin C as a function of demographic, substance use, diagnostic, or micronutrient data using Bayesian log-normal regression. Bayesian regression is seeing increasing use in medicine as a more intuitive means of statistical analysis with many favorable properties (see [46] for an accessible introduction). We used a log-normal distribution since vitamin C levels are naturally lower bounded at 0 (i.e., negative vitamin C levels are not possible) and the empiric distribution has a long right tail (Figure 1). Since Bayesian regressions need to specify prior distributions for parameters, we used weakly-informative priors for coefficients and standard deviation parameters. These weakly-informative priors posit that large effect sizes are unlikely, shrinking estimates closer to 0. This has the favorable benefit of requiring that effects be more robust before claiming significance. Priors were N(0, 1) for coefficients, including intercept terms, and Cauchy⁺(0, 5) for the standard deviation parameter.

Figure 1: Serum vitamin C levels.

Vitamin C levels were collected on n = 221 patients. The overall mean ± standard deviation was 0.68 ± 0.41 mg/dl. A total of 141 patients (64%, 95% confidence interval 57–70%) met criteria for mild vitamin C deficiency (< 0.75 mg/dl). The thin black line marks 0.75 mg/dl.

We report Bayes factors comparing our models to null intercept-only models. Bayes factors are indices that provide relative evidence for one hypothesis over another. They have the advantage over p-values of allowing us to make statements about the null hypothesis, and to quantify whether an analysis is indeterminate [47]. Under common convention, Bayes factors from 3 – 10 provide moderate, 10 – 30 provide strong, 30 – 100 provide very strong, and > 100 provide extremely strong evidence in favor of the hypothesized model, over the null. Bayes factors from 1/3 – 3 are indeterminate, meaning more data are required to make a conclusion. Bayes factors < 1/3 provide evidence for the null model. These boundaries are not strict and, like all statistical inference, conclusions should be interpreted in context [48]. Bayes factors in this manuscript can be interpreted as a statement about parsimony. If we compare a model with many parameters (e.g., demographic-based predictors) to one with a single parameter (intercept-only null) and find the Bayes factor prefers the null, then this means the null model provides a more parsimonious explanation of the data. We also provide Bayes factors for coefficients within a model, which quantifies how much the data provide evidence for a non-zero coefficient.

Since vitamin C data were left-censored, with the minimum reported as < 0.1 mg/dl, we applied left-censoring techniques in all regressions. For the demographic and substance use regression, continuous variables (here, only Area Deprivation Index) were z-scored. For micronutrient regressions (folate, vitamin B12, vitamin D), the input to the model was the z-scored log of the relevant micronutrient data. Since some of the folate and vitamin B12 data were right-censored, we imputed these values before running the regression. For folate, the maximum level was reported as > 20 ng/ml. To impute these right-censored values, we fit a right-censored log-normal distribution to the data, calculated the mean and standard deviation of this empiric distribution, generated random samples from this empiric distribution, and kept those values > 20 to use as inputs into the vitamin C regression model. For vitamin B12, the reported maximum was > 200 pg/ml, with the details of the imputation procedure otherwise the same.

All analyses were performed in R using RStudio. Models were fit using the brms package [49], which is a wrapper for the probablistic programming language Stan [50]. Models were visually inspected with trank plots, which demonstrated adequate chain mixing, suggesting that estimates of the posterior had converged. We computed Bayes factors using the bayestestR package [51, 52].

Results

We identified 221 patients with serum vitamin C levels collected during their inpatient psychiatric admissions. Within this group, 141 patients (64%, 95% confidence interval 57–70%) met criteria for mild vitamin C deficiency (< 0.75 mg/dl), a psychiatrically-relevant category (Figure 1). To better understand what features may predict mild vitamin C deficiency, we identified demographic, substance use, diagnostic, and other laboratory data for each patient.

Demographic and substance use risk factors

Among potential demographic and substance use risk factors (Table 1), univariate analysis revealed that deficient patients were more frequently male (63% vs 46%, p = 0.022) and tended to be older, although this latter effect was not significant (49.2 ± 18.2 years vs 44.5 ± 16.7 years, p = 0.058). We found no differences in race (Deficient: 60% White, 14% Black, 11% Hispanic, 5% Asian, 10% Other; Not Deficient: 64% White, 6% Black, 10% Hispanic, 6% Asian, 14% Other, p = 0.44), current tobacco use (30% vs 38%, p = 0.36), risky alcohol use (AUDIT-C score 1.74 ± 3.33 vs 1.69 ± 3.14, p = 0.65), stable housing (81% vs 76%, p = 0.52), or Area Deprivation Index [41, 42], an index of neighborhood disadvantage relevant for healthcare outcomes (18.2 ± 13.5 percentile vs 21.3 ± 15.4 percentile, p = 0.11).

Table 1:

Demographic and substance use, diagnostic, and laboratory data, split by vitamin C status.

	Total (N=221)	Vitamin C Status		P-value
		Not Deficient (N=80)	Deficient (N=141)
− Demographic and Substance Use Data -
Age (years)	47.5 (±17.8)	44.5 (±16.7)	49 2 (±18.2)	0.063
Sex (female)	35 (43 %)	43 (54 %)	52 (37 %)	0.022
Race				0.44
White	136 (62 %)	51 (64 %)	85 (60 %)
Black	25(11 %)	5 (6 %)	20(14 %)
Hispanic	23 (10 %)	8 (10 %)	15 (11 %)
Asian	12 (5 %)	5 (6 %)	7 (5 %)
Other	25 (11 %)	11 (14 %)	14 (10 %)
Tobacco Use	73 (33 %)	30 (38 %)	43 (30 %)	0.36
Alcohol (Audit-C)	1.72 (±3.26)	1.69 (±3.14)	1.74 (±3.33)	0.65
Housed	175 (79 %)	61 (76 %)	114 (81 %)	0.52
Area Deprivation Index	19.3 (± 14.3)	21.3 (± 15.4)	18.2 (± 13.5)	0.11
− Diagnostic Data -
Depressive	89 (40 %)	35 (44 %)	54 (38 %)	0.51
Bipolar	9 (4 %)	4 (5 %)	5 (4 %)	0.86
Psychotic	85 (38 %)	27 (34 %)	58 (41 %)	0.35
Anxiety	52 (24 %)	22 (28 %)	30 (21 %)	0.38
Substance Use	59 (27 %)	25 (31 %)	34 (24 %)	0.32
Catatonia	31 (14 %)	10 (12 %)	21 (15 %)	0.77
Neurocognitive	37 (17 %)	11 (14 %)	26 (18 %)	0.48
Autism Spectrum	4 (2 %)	3 (4 %)	1 (1 %)	0.27
− Laboratory Data -
Folate (ng/ml)	13.9 (± 5.05)	15 3 (± 4.78)	13.2 (± 5.07)	0.012
Vitamin B12 (pg/ml)	700 (± 374)	699 (± 365)	700 (± 381)	0.98
Vitamin D(ng/ml)	24.1 (± 14.0)	27.3 (± 13.7)	22.3 (± 13.9)	0.0069

We next performed a multivariate analysis of these demographic and substance use factors, since perhaps a combination of factors may better predict vitamin C deficiency. We fit a Bayesian log-normal regression, which predicted vitamin C as a function of demographic and substance use factors. We used Bayes factors to perform model comparison (see Methods for a rationale). Briefly, Bayes factors provide relative evidence for one hypothesis over another and under usual conventions, Bayes factors > 3 favor the experimental hypothesis, and Bayes factors < 1/3 favor the null hypothesis. This model had a Bayes factor of 2.5 × 10⁻⁷ compared to a null intercept-only model, demonstrating that the null model was strongly favored as a more parsimonious explanation of the data. To assess whether any single coefficient may be reasonable predictive, we computed Bayes factors on each coefficient - which quantifies whether the evidence supports a non-zero coefficient - and found that no demographic or substance use factor had a Bayes factor > 0.7. We therefore did not identify a single demographic or substance use predictor that was reasonably predictive of vitamin C levels. To assess this another way, we computed 95% highest density posterior intervals for each coefficient and found that none excluded zero (Figure 2A). In sum, we found no evidence that demographic or substance use data predicted vitamin C levels.

Figure 2: Multivariate regression coefficients.

We fit Bayesian log-normal regressions predicting serum vitamin C as a function of demographic and substance use, diagnostic, or micronutrient factors. We computed the 95% highest density posterior interval for each coefficient (or difference in coefficients, if more interpretable).

A: For the demographic and substance use regression, none of the intervals excluded 0.

B: For the diagnostic regression, none of the intervals excluded 0.

C: For the micronutrient regression, the intervals for folate and vitamin D excluded 0. Mean [95% interval]: folate 0.18 [0.07 – 0.28], vitamin B12 0.09 [−0.01 – 0.19], vitamin D 0.22 [0.12 – 0.31].

Diagnostic risk factors

We next asked whether diagnostic data may predict vitamin C levels. We categorized admissions as related to the following DSM-5-TR diagnoses: depressive, bipolar, psychotic, anxiety, substance use, catatonia, neurocognitive, and autism spectrum. Univariate analyses revealed that no diagnostic label was predictive of vitamin C deficiency (Table 1). Like before, we performed a multivariate analysis and found this model had a Bayes factor of 2.6 × 10⁻⁶, strongly disfavoring diagnoses as predictive of vitamin C. We computed Bayes factors on each coefficient and found that no individual diagnosis had a Bayes factor > 0.9, and that none of the 95% highest density posterior intervals excluded 0 (Figure 2B). We therefore found no evidence that diagnoses were predictive of vitamin C levels.

Micronutrient risk factors

Finally, we explored other markers of micronutrient status. We chose to look at folate, vitamin D, and vitamin B12, as these were routinely collected on admitted patients. Univariate analysis revealed that folate and vitamin D levels were significantly associated with vitamin C status (Table 1). Patients with mild vitamin C deficiency had lower folate levels (13.2 ± 5.1 ng/ml vs 15.3 ± 4.8 ng/ml, p = 0.012) and vitamin D levels (22.3 ± 13.9 ng/ml vs 27.3 ± 13.7 ng/ml, p = 0.0069), but not vitamin B12 levels (700 ± 381 pg/ml vs 699 ± 365 pg/ml, p = 0.98).

We fit Bayesian log-normal regressions to predict vitamin C as a function of folate, vitamin B12, or vitamin D. Here, we found moderate evidence that folate (Bayes factor 9.6), no evidence that vitamin B12 (Bayes factor 0.3), and very strong evidence that vitamin D (Bayes factor 950) predicted vitamin C. The 95% highest density posterior intervals supported this view, with intervals for folate and vitamin D excluding 0 (Figure 2C).

The positive correlation between multiple micronutrients and vitamin C suggests that these deficiencies are most likely a function of relative malnutrition. However, prior work has demonstrated there may be a specific sensitivity to vitamin C deficiency [29]. We therefore sought to address whether the prevalence of vitamin C deficiency would remain high even when folate/vitamin D were sufficiently replete (Figure 3). For each model, we simulated vitamin C levels as a function of folate or vitamin D, across a range of reasonable clinically-relevant values (for the folate model, 2.0 ng/ml to 20.0 ng/ml; for the vitamin D model, 5.0 ng/ml to 50.0 ng/mL). When folate or vitamin D levels were frankly deficient (folate = 2.0 ng/ml, vitamin D = 5.0 ng/ml), predicted vitamin C deficiency was high, at ~ 90%. However, when folate or vitamin D levels were replete (folate = 20.0 ng/ml, vitamin D = 50.0 ng/ml), predicted vitamin C deficiency remained reasonably high (~50–55% even with high levels of folate and vitamin D). Therefore, predicted vitamin C deficiency remained high, even when other micronutrient markers were sufficiently replete.

Figure 3: Predicted vitamin C as a function of folate or vitamin D.

We fit Bayesian log-normal regressions predicting vitamin C as a function of folate or vitamin D. We then simulated data by providing fictive folate or vitamin D levels as inputs to this model to predict vitamin C distributions. For each model, we ran across a range of clinically-relevant values.

A: Left: vitamin C as a function of folate, with a best-fit line ± 95% highest density posterior interval for the mean (in gray shade). Note that the line for the mean is shifted slightly towards the tail (here, upwards) because this is a log-normal distribution, and the mean is closer to the tail than the median. Middle: predicted vitamin C distribution as a function of low (2.0 ng/ml) and high (20.0 ng/ml) levels of folate. The black line marks 0.75 mg/dl, criteria for mild vitamin C deficiency. Right: Predicted probability of vitamin C deficiency (percent of distribution below 0.75 mg/dl) across a range of folate values.

B: Same as above, for vitamin D.

Discussion

Mild vitamin C deficiency is associated with a number of psychiatrically-relevant symptoms, including fatigue, dysthymia, and apathy. Despite its relevance to psychiatry, vitamin C has received comparatively little attention in the literature. Here we find that the prevalence of mild vitamin C deficiency in the inpatient psychiatric setting is remarkably high, at 64% (95% confidence interval 57–70%).

What accounts for this high prevalence? For one, it is remarkably easy to become vitamin C deficient, as the body stores last about 2–3 months [53]. Maintaining this store requires a diet containing fresh produce, since vitamin C is quickly lost when food is cooked [54]. Approximately 20% of US adults consume less than 60mg of vitamin C daily, less than the recommended 75–90 mg/day [9, 55, 56]. Psychiatrically-ill patients are known to have poor diets low in fruits and vegetables [57] and those with psychiatric disorders tend to be socioeconomically disadvantaged [58]. We also know the disease processes themselves contribute. For example, one of the criteria of major depressive disorder is a decrease in appetite, which would certainly contribute to deficiency [45]. We note that the high prevalence of vitamin C deficiency is not unique to populations with psychiatric illness. Hospitalized medical patients show similar rates of deficiency (56–85%), which likely speaks to the multifactorial nature of deficiency [25, 26, 27, 30]. Because vitamin C levels were collected at the beginning of hospitalization, deficiency reflects pre-hospitalization practices, and not differences between the practice patterns of psychiatric vs medical inpatient units.

Our analysis relating vitamin C to folate/vitamin D revealed that the probability of vitamin C deficiency remained high, even when folate/vitamin D levels were sufficiently high. This highlights the importance of independently assessing vitamin C levels, rather than extrapolating vitamin C status from the levels of other micronutrients. One should keep in mind that our analysis relating vitamin C to other micronutrients is a simulation only, which we pursued since our relatively small sample size limited our ability to query vitamin C levels at the extremes of other micronutrient levels. One observation that increases confidence in our analysis is that a sizeable fraction of patients with high levels of folate and vitamin D have vitamin C levels below 0.75 mg/dl (Figure 3).

We found that folate and vitamin D, but not vitamin B12, were associated with vitamin C deficiency. Prior work has demonstrated that not all micronutrients are correlated with vitamin C. One study found that malnourished patients were deficient in vitamin C, but not vitamin A, vitamin E, or vitamin B2, suggesting a particularly sensitivity to vitamin C [29]. In our study, the relationship between folate and vitamin C is likely because the two are found in similar sources, and dietary deficiency would similarly deplete both. The relationship for vitamin B12 and vitamin D is less clear, as natural and fortified versions both are found in similar food sources (predominantly meat and dairy). Perhaps because vitamin D deficiency is more common than vitamin B12 deficiency, there is a natural floor effect with vitamin B12 [59, 60]. More likely, we were simply underpowered to detect a difference with vitamin B12, since a Bayes factor of 0.3 is not definitively in favor of the null hypothesis. In this case, malnutrition would be the most straightforward explanation.

Prior literature has demonstrated that male sex, smoking, and low socioeconomic status are commonly-found risk factors for vitamin C deficiency [19]. We only found age as a risk factor on univariate analysis, which disappeared with multivariate analysis. We believe this is because vitamin C deficiency is particularly prevalent in our population, minimizing or negating the effect of additional risk factors which may be relevant for the general population. We refer the interested reader to [61] for a thoughtful contemporary review on the risk factors underlying vitamin C deficiency.

Our study has a number of limitations. Unlike other micronutrients (vitamin B12, vitamin D, and folate) which were routinely collected, vitamin C levels were collected in only a subset of patients. In general, vitamin C was collected if there was a concern for a produce-limited diet, if patients were unable to provide a reliable history (e.g., psychosis), or based on clinical judgment (e.g., as part of a diagnostic workup, physical exam findings concerning for scurvy, etc.). This bias in sampling might explain, in part, the high rates of vitamin C deficiency we observed - equivalent to rates observed in medically-hospitalized patients. Because the psychiatric symptoms of vitamin C deficiency are relatively mild and since our study was performed in the inpatient setting, where symptoms tend to be more extreme, we found that symptomatology had limited utility in determining who was vitamin C deficient. For example, essentially all patients with depression who require inpatient hospitalization will report dysthymia, fatigue, or apathy, regardless of whether they are vitamin C deficient. Consistent with this, exploratory data analysis revealed no enrichment of symptoms in those with low vitamin C levels. Additionally, if symptomatology were the primary factor driving bias in sampling, one may expect a lower rate of vitamin C deficiency in patients unable to provide a clear history (e.g., psychosis or catatonia), as this patient population won’t necessarily be enriched for symptoms of vitamin C deficiency. As we showed in Figure 2B, diagnostic factors were not predictive of serum vitamin C. We cannot rule out that this inference is limited by small sample size, since most Bayes factors for individual diagnostic categories were indeterminate, suggesting the need for more data. We hypothesize that the relationship between vitamin C and symptomatology will be more pronounced in the outpatient setting, where milder psychiatric symptoms are more prevalent.

Our study has a number of other limitations. We used records from one inpatient psychiatric unit in an academic hospital serving a large metropolitan area. Unique practices within our hospital system likely affected data collection practices - for example, there is a greater consideration for the psychoactive impact of nutrition than is standard for the field of psychiatry. Other limitations include using Zip code, a fairly crude measure, as a proxy for socioeconomic status, as there can be tremendous variability in socioeconomic status within a single Zip code. Finally, we likely had limited power in identifying various risk factors, a number of which are firmly established in the literature (e.g., male sex, cigarette smoking). Overall, these limitations make it clear that a prospective study, in which vitamin C is collected in a more unbiased manner, is a necessary next step to better understand the prevalence of vitamin C deficiency and associated risk factors in the inpatient psychiatric setting.

There is a physiological basis for why mild vitamin C deficiency may predispose to psychiatric symptoms. The cerebrospinal fluid vitamin C concentration is 3–5 fold higher than plasma concentrations, suggesting higher metabolism by the nervous system [62]. Vitamin C is involved in norepinephrine synthesis, modulation of neurotransmission by dopamine and glutamate, and the regulation of catecholamine and acetylcholine release from synapses [16]. It is also a powerful antioxidant, protecting against excitotoxicity from glutamate [16]. Vitamin C deficiency additionally disrupts cortico-basal-ganglia processing, which may predispose to apathy and difficulty enacting willed motor movements [63].

The psychiatric consequences of vitamin C deficiency generally correct with adequate repletion. Notably, repletion improves mild mood symptoms across a range of settings [26, 27, 64], including healthy individuals [65]. In contrast, there seems to be no such improvement in major depressive disorder based on a recent systematic review [66]. Notably, however, this review did not assess vitamin C status in depressed patients prior to supplementation, which is a frequent limitation of nutrition intervention studies that often yield negative findings. While it remains unlikely (yet unclear) that vitamin C deficiency contributes to profound neurovegetative depression, it can likely limit recovery from acute psychiatric crises. Given the high prevalence of vitamin C deficiency in the inpatient setting combined with the difficulty of identifying deficient patients based on risk factors, we make the following suggestions for changing clinical practice: we suggest routinely collecting vitamin C levels to allow for patient-specific recommendations; we recommend psychiatric treatment units ensure patient diets contain daily fresh fruits and vegetables (and to initiate such diets after the serum level is drawn). For patients with limited oral intake or with differing dietary preferences, oral repletion (500mg BID or renal-attenuated dosing) is a cheap and well-tolerated alternative.

Nutrition remains an important and often overlooked aspect of psychiatric illness. The field is in need of well-designed studies that identify patients with specific deficiencies and measure clinical outcomes pre- and post-repletion. Such work is a key foundation of precision psychiatry and will allow us to treat patients in a targeted manner.

Clinical Points.

• Although it is well-known that mild vitamin C deficiency results in psychiatrically-relevant symptoms, the prevalence of vitamin C deficiency in the inpatient psychiatric setting has not previously been reported in the literature.

• Mild vitamin C deficiency is remarkably common in the inpatient psychiatric setting, with 64% of patients meeting criteria.

• It is worth obtaining vitamin C levels as part of admission screening labs and ensuring that inpatients have access to fruit- and vegetable-rich diets.