The Incidence and Risk Factors for Hypotension After Spinal Anesthesia Induction: An Analysis with Automated Data Collection

Abstract

We sought to identify factors that are associated with hypotension after the induction of spinal anesthesia (SpA) by using an anesthesia information management system. Hypotension was defined as a decrease of mean arterial blood pressure of more than 30% within a 10-min interval, and relevance was defined as a therapeutic intervention with fluids or pressors within 20 min. From January 1, 1997, to August 5, 2000, data sets from 3315 patients receiving SpA were recorded on-line by using the automatic anesthesia record keeping system NarkoData. Hypotension meeting the predefined criteria occurred in 166 (5.4%) patients. Twenty-nine patient-, surgery-, and anesthesia-related variables were studied by using univariate analysis for a possible association with the occurrence of hypotension after SpA. Logistic regression with a forward stepwise algorithm was performed to identify independent variables (P < 0.05). The discriminative power of the logistic regression model was checked with a receiver operating characteristic curve. Calibration was tested with the Hosmer-Lemeshow goodness-of-fit test. The univariate analysis identified the following variables to be associated with hypotension after SpA: age, weight, height, body mass index, amount of plain bupivacaine 0.5% used for SpA, amount of colloid infusion before puncture, chronic alcohol consumption, ASA physical status, history of hypertension, urgency of surgery, surgical department, sensory block height of anesthesia, and frequency of puncture. In the multivariate analysis, independent factors for relevant hypotension after SpA consisted of three patient-related variables (“chronic alcohol consumption,” odds ratio [OR] = 3.05; “history of hypertension,” OR = 2.21; and the metric variable “body mass index,” OR = 1.08) and two anesthesia-related variables (“sensory block height,” OR = 2.32; and “urgency of surgery,” OR = 2.84). The area of 0.68 (95% confidence interval, 0.63–0.72) below the receiver operating characteristic curve was significantly greater than 0.5 (P < 0.01). The goodness-of-fit test showed a good calibration of the model (H = 4.3, df = 7, P = 0.7; C = 7.3, df = 8, P = 0.51). This study contributes to the identification of patients with a high risk for hypotension after SpA induction, with the risk increasing two- or threefold with each additional risk factor.

Hypotension, with an incidence of 15%⁽¹⁾ to 33%⁽²⁾, is one of the most frequent side effects of spinal anesthesia (SpA). The clinical importance of this side effect was shown in a study by Sanborn et al. ⁽³⁾, who proved that hypotensive episodes detected by an automated record-keeping system clearly correlate with mortality.

Different incidences of hypotension, as reported in the literature, can be due to varying definitions and different methods of measurement. In most studies, blood pressure readings were documented manually. Nonetheless, various authors have shown that automatic gathering of vital variables on-line, along with an exact definition of hypotension, can lead to more precise and comprehensive documentation of adverse events when compared with manual documentation. This also applies for intraoperative hypotension ^(3–5).

Many studies have been conducted concerning prophylaxis and therapy of hypotension after SpA or epidural anesthesia. Predictions of these events, however, have been addressed by only a few authors for SpA ^(1,2) and epidural anesthesia ⁽⁶⁾. The goal of this study was to identify factors associated with hypotension after SpA, by use of data from computerized anesthesia records of a large patient collective.

Methods

From January 1, 1997, to August 5, 2000, data sets of 3315 patients with SpA were recorded on-line with the anesthesia record keeping system NarkoData (IMESO GmbH, Hüttenberg, Germany) ⁽⁷⁾. The software collects all data relevant to anesthesia during the procedure, including biometrical data, administered drugs, laboratory results, vital data, and the data set for quality assurance according to the German Society of Anesthesiology and Intensive Care Medicine ⁽⁸⁾.

All patient-related data gathered during the preoperative ward round, informed consent of the patient, results of clinical examination, and additional investigations were recorded by the anesthesiologist in the electronic anesthesia record. The program guides the anesthesiologist through the NarkoData menu with mandatory data entry into compulsory fields. For evaluation purposes, missing variables were classified as “not pathological,” assuming that these would have otherwise been assessed.

Mean arterial blood pressure (MAP) was recorded at least every 5 min with noninvasive measurement and every 3 min with invasive measurement. Record files were imported into the database (Oracle 7^®; Oracle Corp., Redwood Shores, CA) after they were run through plausibility and integrity checks. These checks included evaluation of biometrical data, preoperative laboratory results, medical history, diagnoses, procedures, staff involved, and mandatory time periods, to name a few. The Voyant^® software (Brossco Systems, Espoo, Finland) was used for database queries and evaluation.

Structure Query Language statements were defined for automatic detection of episodes of hypotension occurring within 30 min after lumbar puncture. For this study, we investigated only cases of relevant (requiring therapeutic intervention) hypotension. Episodes of relevant hypotension were defined as follows: a decrease in MAP of more than 30% from baseline values within a 10-min interval, plus a therapeutic intervention by the attending anesthesiologist (the administration of an additional bolus of 500 mL of crystalloid and/or colloid and/or the administration of a vasoconstrictor—amezinium metilsulfate [Supratonin^®, Grünenthal GmbH, Aachen, Germany], cafedrine/theodrenaline [Akrinor^®, ASTA Medica AWD GmbH, Frankfurt, Germany], epinephrine, or norepinephrine—within 20 min of onset of decrease in MAP).

Adherence to this strict definition allowed a detection of hypotension with high specificity and low effect of artifacts. (For example, 3 min after the induction of SpA, MAP is 100 mm Hg and decreases to 65 mm Hg within 8 min. Four minutes later, the anesthesiologist administers a vasoconstrictor. This constellation would have been detected as an event of relevant hypotension.)

Electronically detected events were checked for artifacts by two independent investigators who controlled a randomly selected 20% of all anesthetic records containing the event “relevant hypotension.” The predictive power of 13 patient-related, 4 operative, and 12 anesthesia-related variables (8 variables pertaining to regional anesthesia) was studied:

• 1. Patient-related variables: age, height, weight, body mass index (BMI), sex, ASA physical status ⁽⁹⁾, active cigarette use, chronic alcohol consumption (defined as more than three alcoholic drinks per day), chronic heart failure (New York Heart Association classification I –IV), preoperative history of hypertension or hypotension, vascular diseases, endocrine diseases, and preoperative antihypertensive treatment (angiotensin-converting enzyme inhibitors, β-adrenergic blockers, calcium antagonists, diuretics).
• 2. Surgical variables: admission status (inpatient/outpatient), emergency or elective surgery, surgical department (orthopedic and trauma surgery, general surgery, urology, gynecology, and others), and type of surgical procedure according to the International Classification of Procedures in Medicine.
• 3. Anesthetic variables: oral premedication with 3.8 or 7.5 mg of midazolam (yes/no), amount of IV crystalloid/colloid preload administered before SpA, IV sedation after SpA (midazolam, propofol, or both), and time interval between SpA puncture and start of operation.
• 4. SpA variables: spinal needle (Atraucan, Quincke, Whitacre, Sprotte), spinal needle size (22- to 29-gauge), site of SpA puncture (L1-2 to L5-S1), frequency of puncture (1 to 4, ≥5), type and dose of local anesthetic (plain bupivacaine 0.5% or hyperbaric mepivacaine 4%), sensory block height measured 10 min after application of the block by thermal stimulation with cold alcohol spray, and local complications after SpA puncture (bleeding, paresthesia).

There were no cases of SpA in obstetric surgery. Two-hundred-seventeen patients receiving combined anesthesia had to be excluded from the study (104 for additional peripheral regional anesthesia and 113 for intubation procedures). A total of 3098 patients were included in this study.

Data analysis and statistics were performed by using the statistics program SPSS^® (SPSS Software GmbH, Muenchen, Germany). The dichotomous variable “relevant hypotension after SpA (yes/no)” was used as a target criterion.

First, variables were checked with univariate analysis for associations with a relevant decrease of MAP. We calculated mean value, sd, median, interquartile range, and 95% confidence interval as metric variables. Metric variables were compared by using the nonparametric Mann-Whitney U-test. Categorical variables were assessed for a significant association with “relevant hypotension” by using either χ² statistics or Fisher’s exact test.

Second, logistic regression was used to investigate independent factors with a significant association to “relevant hypotension” within a multivariate model. A forward stepwise algorithm (inclusion criteria: log likelihood test ratio based on maximum likelihood function) was used. At each step, independent variables not yet included in the equation were tested for possible inclusion. The variable with the strongest significant contribution (P < 0.05) to improving the model was included. Variables already included in the logistic regression equation were tested for exclusion on the basis of the probability of a log likelihood test ratio. The analysis ended when no further variables for inclusion or exclusion were available. Variables in the logistic regression equation were indicated as “1” if present and “0” if absent. Continuous values were calculated as absolute values. Furthermore, logistic regression was used to estimate the coefficients (β) of these variables. On the basis of the results, the probability of the event “relevant hypotension” may be estimated with the logistics function.

The model’s discriminative power was tested by cross-validation by using the “leaving one out” technique ⁽¹⁰⁾. In this type of analysis, one data set is retained for validation and not used for creating the model. For this data set, the probability of the observed event (relevant hypotension) is calculated according to the equation. The statistical significance of this technique is enhanced by using each data set as a control for validation. In our study, this resulted in 3098 control data sets validating the model. To this end, the probability of each database case was calculated and a receiver operating characteristic (ROC) curve created. The ROC curve plots the percentage of true-positive values (sensitivity) on the basis of the individual score against the percentage of false-positive values (1 − specificity). The area under the curve indicates the accuracy of the calculated model between 0 and 1. This may be interpreted as the probability of correct patient classification in one of the two categories (“relevant hypotension” yes/no). An area of 0.5 indicates that the predictive accuracy equates a random selection.

We used the Hosmer-Lemeshow goodness-of-fit H and C statistics to evaluate the overall calibration of the model equation by testing the null hypothesis that the mean predicted and observed incidences of relevant hypotension were equal ^(11,12). For this reason, the hypothesis has to be retained for confirmation of the model’s fitness. The level of significance should meet a minimum of P > 0.05 and preferably P > 0.2 to allow an indirect control of the β failure. The observed and expected incidences are graphically shown in calibration curves with the number of patients per group.

Results

Altogether, a decrease in MAP within 30 min after the induction of SpA was recorded in 3074 (99.2%) of 3098 patients. In 46.8% (n = 1450), MAP decreased by 10% to 20%, and therapeutic intervention was recorded in 52.9% (n = 767) of this group. In 19.8% (n = 613) of all cases, there was a decrease in MAP of 20% to 30%; 50.4% (n = 309) of these patients received a therapeutic intervention. In 8.2% (n = 254) of all cases, MAP decreased more than 30%. An episode of relevant hypotension—a decrease of MAP of more than 30% from baseline values within a 10-min interval in the first 30 min after SpA induction, requiring therapeutic intervention until 20 min after the onset of the decrease—was detected in 5.4% (n = 166) of all cases. These patients (n = 166) with relevant hypotension were included in the analysis. The valuation of samples for accuracy determination revealed no artifacts among automatically detected events.

The following variables were identified with univariate analysis as having an association with a higher incidence of hypotension: age, weight, height, BMI, amount of colloid infusion given before puncture, chronic alcohol consumption, ASA physical status, preoperative history of hypertension, long-term antihypertensive therapy, urgency of surgery, operative department, sensory block height of anesthesia 10 min after application of the local anesthetic, and frequency of puncture. Mean value and sd, and median with the 25th and 75th percentiles of the distribution of metric variables, are given in Table 1. Table 2 shows the distribution of ordinal and nominal variables and their relation to the incidence of relevant hypotension.

Table 1: Univariate Analysis (Mann-Whitney U-test) of the Metric Predictors for Hypotension with Mean Values and sd or Median with the 25th and 75th Percentiles of the distribution

Table 2: Univariate Analysis of the Association of Variables on Relevant Hypotension (χ² test or Fisher’s exact test)

The following factors did not show significant association: type of premedication, sedation after application of SpA, operative procedure, and time interval between block and start of surgery (Table 3). In contrast, there was a highly significant relationship (P < 0.01) for the surgical department involved. Relevant hypotension occurred approximately twice as often in general surgery and gynecology as compared with trauma surgery and urology. Table 4 shows the correlation between regional anesthesia variables and relevant hypotension.

Table 3: Univariate Analysis of the Association of Variables on Relevant Hypotension (χ² test or Fisher’s exact test)

Table 4: Univariate Analysis of the Association of Regional Anesthesia-Related Variables with Relevant Hypotension (χ² test or Fisher’s exact test)

The results of stepwise logistic regression analysis are summarized in Table 5. Among patient-related factors, chronic alcohol consumption showed the strongest association (odds ratio [OR] = 3.05), indicating a more than threefold increased risk for this side effect. Preoperative history of hypertension (OR = 2.21) and BMI (OR = 1.08) also represent patient-related factors associated with increased risk for hypotension. Within the logistic regression model, the risk increased with the level of analgesia (sensory block height after 10 min classified in <T6 and ≥T6, OR = 2.32). Emergency procedures provided a nearly threefold risk for relevant hypotension (OR = 2.84). Because of a lack of significant association within the model, all other indicators were excluded from the logistic regression model.

Table 5: Results of Logistic Regression Analysis

The probability of a relevant hypotension could be calculated by the following logistic equation:MATH

The factors for “chronic alcohol consumption,” “preoperative history of hypertension,” and “ur- gency” were indicated as “1” if present and “0” if absent. For “block height,” the factor “1” equates to a level of analgesia ≥T6 and “0” to a level of analgesia <T6. The BMI was treated as a continuous variable in the logistic regression analysis. Therefore, the OR for BMI corresponds to that associated with a 1 kg/m² increase in BMI.

Probabilities obtained after validation of the score by “leaving one out” were used for calculation of the ROC curve (Fig. 1). The area under the ROC curve, defined as the measure of accuracy, was 0.68. The 95% confidence interval (0.63–0.72) is represented by area.

Figure 1: Receiver operating characteristic (ROC) curve of the logistic regression model (inclusion criterion). The area under the ROC curve was 0.68, with a 95% confidence interval of 0.63–0.72.

A probability of 0.10 resulted in a sensitivity of 0.29 and a specificity of 0.92. This threshold of 0.10 showed the best relationship between sensitivity and specificity for factor inclusion. The percentage of correctly categorized patients on the basis of this threshold was 89%. Other values and their respective variables can be used: sensitivity, specificity, and the total number of correctly categorized patients on the basis of other limits. These can be ascertained by using the ROC curve (Fig. 1).

The Hosmer-Lemeshow statistics showed, in the goodness-of-fit H statistic, H = 4.3, df = 7, and P = 0.7 and, in the goodness-of-fit C statistic, C = 7.3, df = 8, and P = 0.51, demonstrating good calibration of the model. The model equation shows good correspondence between observed and expected values within these groups. The calibration curve shows the observed and expected “relative hypotension” together with the number of categorized patients (bar diagram) depending on the probability of hypotension (Fig. 2).

Figure 2: Comparison of predicted (—♦—) and observed (—▪—) incidence of hypotension after spinal anesthesia. Estimated risk of hypotension, observed hypotension, and the corresponding number of patients (columns).

Discussion

Hypotension is one of the most frequent side effects of SpA. In this study, we found a decrease of MAP in all of our patients within 30 minutes after the induction of SpA. A decrease in MAP of more than 30% after SpA induction occurred in 8.2% of the studied patients. In 5.4% of all patients, MAP decreased more than 30%, followed by therapeutic intervention by the attending anesthesiologist. These events were defined as relevant hypotension. With the help of automated recorded data, we could detect five independent variables having an association with hypotension after SpA induction. The risk increased two- or threefold with each additional risk factor. The knowledge of these risk factors should help clinicians treat hypotension more effectively or to use alternative methods of SpA.

Carpenter et al. ⁽²⁾ described hypotension with an incidence of 33% in their study. They defined hypotension as systolic blood pressure <90 mm Hg or, alternatively, as a 10% decrease from the baseline in patients with baseline blood pressure <90 mm Hg. Tarkkila and Isola ⁽¹⁾ defined hypotension as a decrease in systolic blood pressure of more than 30% from the preanesthetic value or a decrease of systolic blood pressure to less than 85 mm Hg. They detected episodes of hypotension in 15.3% of their patients. The relatively small incidence of relevant hypotension observed in our study can be explained by the strict definition. As already explained previously, we chose this strict definition to detect episodes of hypotension with high specificity and at the same time with a low effect of artifacts. However, considering hypotensive episodes with a decrease of MAP of more than 10% or 20%, the incidence in our study is comparable to results of other studies ^(1,2). The definitions of hypotension used in the studies cited previously are questionable because the authors define hypotension exceeding a lowest boundary or in choosing the first blood pressure reading as baseline. However, this approach does not take individual patient processes into account. Contrary to these studies, our definition of hypotensive episodes is based on MAP instead of systolic blood pressure readings, because MAP is the most important blood pressure variable concerning organ perfusion.

Additionally, blood pressure readings and clinical results were all recorded automatically in our study. In multiple studies, authors using patient data management systems in risk management and quality assurance have repeatedly proven a benefit in defining and detecting adverse events recorded automatically compared with those recorded manually ^(3–5).

The correlation of circulatory instability with higher cephalic levels of neuraxial blockade has already been proven in previous studies ⁽²⁾. Circulatory regulation is affected by a blockade of the sympathetic nervous system, with resulting reductions in both venous return and systemic vascular resistance. Furthermore, when the level of analgesia exceeds T4, cardioacceleratory fibers are blocked, leading to a decrease in heart rate and cardiac output.

Patients with chronic alcohol consumption showed a threefold increased risk for a relevant decrease in blood pressure. Because of alcohol-based neuropathy, the sympathetic nervous system is affected, leading to orthostatic dysregulation ⁽¹³⁾. Because lower levels of catecholamines can be observed after alcohol deprivation ⁽¹⁴⁾, alcoholics seem to compensate for latent hypovolemia with an increased output of catecholamine. SpA reveals this compensatory mechanism.

Similar observations can be made in emergency patients. Patients undergoing elective surgery are adequately examined and receive preoperative therapy of existing diseases and volume states. Emergency patients, in contrast, are not always in homeostasis. Preexisting conditions, especially cardiac diseases, may be inadequately assessed. Furthermore, stress encountered in emergency situations, which leads to an increased sympathetic tone, or trauma-associated stress and blood loss are factors that can affect circulatory variables, all which must be considered during SpA.

In patients with known hypertension, the risk for a relevant decrease in blood pressure is nearly twofold. This association has already been evaluated by Racle et al. ⁽¹⁵⁾.

Estimating the amount of local anesthesia needed for sufficient blockade is more difficult in obese patients ^(16,17). A possible explanation might be an increased abdominal pressure with compression of the subarachnoid cavity and a reduction of the cerebrospinal fluid. Nevertheless, an expected reduction of local anesthetic amount in patients with increased BMI could not be observed. However, it appears that the high risk for relevant hypotension in obese patients does not coincide with an expected increased level of anesthesia. The extent of sensory block correlates with lumbosacral cerebrospinal fluid volume and not with body habitus ⁽¹⁸⁾.

There is a controversial discussion concerning the possible protective effect of the type and amount of infusion before spinal block on hypotension ^(19,20). In contrast to Rout et al. ⁽²¹⁾, who found that prophylactic application of IV fluids did not prevent hypotension, we observed a marginal effect of colloids in the univariate analysis, although this was without relevance in the multivariate analysis.

The model for prediction of hypotension during SpA showed a good correlation of observed and expected frequencies. With P = 0.51 and P = 0.70 in the goodness-of-fit test, the β failure was well controlled, meaning that the model equation showed good calibration. All calculated probabilities in the model were low because investigated predictors, except BMI, were indicator variables. Patients who did not fulfill any indicator condition consequently had a low probability. Because hypotension was detected in only 165 patients, high probabilities were rare or remained vacant.

The ROC curve was used to summarize the findings of the multivariate analysis. The area under the ROC curve, acting as the measure of accuracy, was 0.68, indicating a moderate discriminative power for the final model. Comparison of a validation sample with data from another medical center that used the same computerized anesthesia record system could have been helpful to improve validating the importance of the identified risk factors.

Although we were able in this exploratory investigation to demonstrate that data collected with an anesthesia information management system are suitable for developing a multivariate model for identifying risk factors for relevant hypotension, this must be validated in further studies. Some limitations of this study must be acknowledged. A retrospective analysis of routine medical data collected on-line cannot be as objective as a prospective study with complete and uniform data. This is especially true for data recorded as “not documented” and thus regarded as “not pathological” for the purposes of this study, a fact that has to be retained for interpretation of data and study results.

The short observation period of 30 minutes after SpA induction, chosen to exclude surgery-related causes for hypotension, leads to a decreased overall incidence of hypotension. Measuring the sensory level 10 minutes after induction does not necessarily meet the peak sensory block height. In clinical routine, evaluating the extent of sensory block is not performed every five minutes, similar to prospective studies ⁽²²⁾. Thus, the maximum extension may be missed. Evaluation after 10 minutes seems sufficient for routine purposes.

Finally, it may be said that patients developing relevant hypotension during SpA will probably also tend to develop hypotension during general anesthesia. Therefore, the anesthesiologist should not necessarily refrain from using SpA in patients with independent risk factors for hypotension. However, the knowledge of these risk factors should be useful in increasing vigilance in those patients most at risk for hypotension, in allowing for more timely therapeutic intervention, or even in suggesting the use of alternative methods of SpA, such as titrated continuous or small-dose SpA.

We would like to thank the Büro für Statistik GmbH for the help in data management and statistical evaluation.